A Theory of Security Strategy for Our Time Defensive Realism

A Theory of Secur ity Str ategy for Our Time

This page intentionally left blank

A Theory of Security Str ategy for Our Time

Defensive Realism

Shiping Tang

A THEORY OF SECURITY STRATEGY FOR OUR TIME

Copyright © Shiping Tang, 2010.

All rights reserved.

First published in 2010 byPALGRAVE MACMILLAN®in the United States—a division of St. Martin’s Press LLC,175 Fifth Avenue, New York, NY 10010.

Where this book is distributed in the UK, Europe and the rest of the world, this is by Palgrave Macmillan, a division of Macmillan Publishers Limited, registered in England, company number 785998, of Houndmills, Basingstoke, Hampshire RG21 6XS.

Palgrave Macmillan is the global academic imprint of the above companies and has companies and representatives throughout the world.

Palgrave® and Macmillan® are registered trademarks in the United States, the United Kingdom, Europe and other countries.

ISBN: 978–0–230–62313–2

Library of Congress Cataloging-in-Publication Data

Tang, Shiping. A theory of security strategy for our time : defensive realism /

Shiping Tang. p. cm. ISBN 978–0–230–62313–2 (alk. paper) 1. Security, International. 2. Realism—Political aspects.

3. International relations—Philosophy. I. Title.

JZ5588.T367 2010355�.0335—dc22 2009025314

A catalogue record of the book is available from the British Library.

Design by Newgen Imaging Systems (P) Ltd., Chennai, India.

First edition: March 2010

10 9 8 7 6 5 4 3 2 1

Printed in the United States of America.

To Herbert Butterfield, John Herz, and Arnold Wolfers, who founded this more optimistic strain of realism after one of the

bloodiest centuries in human history

To Robert Jervis, who provided the core intellectual foundation


Contents

List of Figures ix

List of Tables xi

Preface xiii

Introduction 1

Part I Preparing the Ground

One Clearing the Theoretical Underbrush 9

Two The Security Dilemma: A Conceptual Analysis 33

Three The Security Dilemma and War Revisited 73

Part II Strategic Choices

Four The Operational Code of Defensive Realism 99

Five Reassurance: A Defensive Realism Theory of Cooperation-building 129

Part III Implications and Conclusions

Six Recategorizing Realism Theories 165

In Lieu of Conclusion: Policy Implications 179

Appendix I: World War I and the Cold War Revisited 185

Appendix II: Liberals, Conservatives, Doves, and Hawks 189

Notes 193

Bibliography 227

Index 243


Figures

2.1 The causal link from anarchy to the security dilemma and war 41

2.2 Conflict of interest and the security dilemma 442.3 The continuum: From a security dilemma to a spiral 654.1 Realism’s ladder of strategies 1044.2 Realism’s ladder of alliance strategies 1055.1 Trust versus mistrust 132


Tables

1.1 Political realism and realisms 111.2 Dichotomies for labeling the two types of states 242.1 The security dilemma according to Butterfield,

Herz, and Jervis 382.2 Common errors about the security dilemma 604.1 State behaviors in war and peace: Offensive realism

versus defensive realism 1114.2 The general operational codes of the three approaches

at a glance 1265.1 Types of response to a reassurance attempt 1385.2 Forms of reassurance attempt 1496.1 Post-Waltz realism theories at different levels of analysis 170


Preface

In 2002, I realized that debates among the three grand theories of inter-national relations or IR—namely, offensive realism, defensive realism, and neoliberalism—cannot be resolved because they have relied on a non- evolutionary approach toward IR and human history in general. Yet the system known as international politics (as part of human society) has always been an evolutionary system. As such, only a social evolutionary approach can understand the evolution of the system adequately, and non-evolutionary theories are inherently incapable for the task.

Since 2002, I have set out to develop a social evolutionary interpreta-tion of the macro-history and grand theories of international politics, which I hope to eventually publish as a separate book, and I have sketched out some of the basic ideas in an article (Tang 2009b). As I began to advance a social evolutionary interpretation of international politics, however, another “unhappy” reality became evident to me. Among the three grand theories of international politics, only John Mearsheimer has bothered to develop something close to a coherent statement for offensive realism. In contrast, both defensive realism and neoliberalism lack such a coherent statement.

This lack of coherent statement poses a major obstacle for my develop-ing a social evolutionary interpretation of international politics. I cannot depict the different stages of international politics with offensive realism, defensive realism, or neoliberalism by simply relying upon existing litera-ture. To advance a social evolutionary interpretation of international poli-tics, I have to first make clear what constitutes offensive realism, defensive realism, and neoliberalism, and only thereafter can I employ them to depict an offensive realism world, a defensive realism world, and a neoliberalism world in international politics. Because my evolutionary interpretation of international politics first dealt with the evolution from a paradise-like anarchy to an offensive realism world and then from an offensive realism world to a defensive realism world, and also because Mearsheimer has pro-vided a decent statement for offensive realism, I have decided to formulate a coherent statement for defensive realism first. This book is the result of such an effort.

xiv / preface

Bob Powell has to be the first individual for me to thank. He was the one most responsible for introducing me to and getting me hooked with those “opaque” theories of international politics. Bob has been extremely demanding on making assumptions explicit, drawing rigorous deduction, and being precise with elaboration. While I am sure that Bob will still be somewhat disappointed with this book, I hope he at least finds some com-fort in the fact that I have been trying hard.

The second individual I need to thank is Bob Jervis. I sent him the first e-mail as I was just beginning to appreciate his enormous stature in the field, and the flow of e-mails between the two of us has never ceased since then. I hope this book will somewhat convince him that he did not waste too much of his precious time on somebody whom he only came to meet after almost ten years of e-mailing. For this book, Bob has provided not only ini-tial encouragement but also detailed comments on chapters 1, 2, and 5.

I next thank Ned Lebow, who has been extremely generous with his time, providing timely insights, criticisms, and advice while finishing his own 762-page book.

The following individuals—some of whom I have yet to meet—have made available their time and provided expertise in answering my queries and commenting on my work over the years. They are: Amitav Acharya, Rajesh Basrur, Emrys Chew, Tom Christensen, Keith Clemenger, Alan Collins, Taylor Fravel, Mike Glosny, Avery Goldstein, Peter Gries, Richard Herrmann, S.P. Harish, Andy Kydd, Christopher Layne, Jeff Legro, Gerald Geunwook Lee, Jon Mercer, Bob Ross, Matthew Rudolph, David Rousseau, Randy Schweller, Alexander Vuving, and Alex Wendt.

Let me also take this opportunity to express my gratitude to the Institute of Asia-Pacific Studies (IAPS), Chinese Academy of Social Sciences, Beijing, and the S. Rajaratnam School of International Studies (RSIS), Nanyang Technological University, Singapore, with special thanks to Director Zhang Yunling of IAPS, and Dean Barry Desker of RSIS. Under their able leadership, these two institutions provided me with a stimulating working environment.

Chong Yee Ming and Lai Foong Yee at RSIS library provided wonder-ful logistical support in finding and borrowing books from various librar-ies, often within very short notice. Beatrice Bieger and Yan Hong provided excellent research assistance during the finishing stage of this book.

My gratitude also goes to four of my special friends: Tao Cui, Hongmei Li, Honglin Li, and Yu Zhang. Their unflinching faith in me has always been an inspiration and an extremely important source of support, encour-agement, and understanding.

A shorter version of chapter 2 appeared in Security Studies, 18 (3). I thank the anonymous reviewers and the two editors of the journal (Michael

preface / xv

Desch and William Wohlforth) for their perceptive criticisms and editing. I also thank Taylor and Francis for permitting the use of the published material.

My deepest gratitude goes to my extended family. My parents not only endowed me with the stamina for toughening it out but also forgave me for spending so little time with them over the years. My wife and my parents-in-law have shouldered most of the burden of raising my little boy (now six years old). And finally I thank my little boy for being so understanding, I have spent so little time in the past couple years with him that when I try to play with him for a few minutes he gets so excited.

This book can only repay part of the debt that I owe all of you.


Introduction

This book advances a coherent statement of defensive realism. By doing so, I seek to serve two purposes. First, I provide a menu of strategic choices for statesmen. Because the world has firmly evolved from an offensive realism world in the past to a defensive realism world in which we live today (Tang 2009b),1 defensive realism is an appropriate guide for security strategy in our time. Indeed, unless all the major powers are firmly “socialized” into defensive realist states, a neoliberalism world—that is, a more rule-based international politics—will not come into existence. For policymakers who are more interested in the strategic choices within defensive realism, they can move directly to chapter 4 and 5, and refer to other chapters when they need further clarifications.2

Second, I add to our understanding of defensive realism as a grand the-ory of International Relations (IR) in particular and our understanding of IR in general. By making explicit some of the underappreciated and unar-ticulated differences between defensive realism and offensive realism (and, to a less extent, neoliberalism too), I hope to facilitate and contribute to the ongoing debates among major paradigms of IR.

Grand Theories and Great Debates in International Politics

A theory is a scientific and logical explanation of social facts. A grand the-ory is a theoretical approach, or a paradigm/research program in Kuhnian and Lakatosian language (Kuhn 1970; Lakatos 1970). For a paradigm to qualify as a grand theory of international politics, it must have a hard core that consists of a coherent set of core assumptions, a protecting belt that are derived from the hard core, and a rich array of mid-level theories that explain more specific social facts. Most critically, the paradigm’s hard core must capture some fundamental features of international politics in his-tory reasonably well, ontologically speaking.

With the above criterion, I consider only three major theoretical approaches to have legitimate claim as grand theories of international pol-itics. The three grand theories are: offensive realism, defensive realism, and neoliberalism. As I argue elsewhere (Tang 2009b), these three grand

2 / theory of security strategy

theories roughly capture a world that we had experienced before 1945 (the offensive realism world), a post-1945 world that we have been experiencing (the defensive realism world), and a world that we may be in the process of making (a neoliberalism world).

In the past century, debates among major grand theories of IR have, to a very large extent, shaped the development of IR as a science. These debates are not just meaningless and egoistic quarrels among scholars. Instead, these debates bear great relevance for the real world because each grand theory carries implicit and explicit messages for states’ security strat-egies. Each grand theory paints a different picture of our past, present, and future, and how we understand our past and present certainly affects how we behave today, and how we behave today will undoubtedly shape our future. There is no doubt that different grand theories have informed the debate on the future of Europe after the cold war (e.g., Mearsheimer 1990; Van Evera 1990–1901), America’s debate about how to cope with a rising China, Europe’s debate about how to live with a post-Soviet Union Russia, China’s debate about how to live with America as the reigning hegemon, and China’s and India’s debate about how to live with each other as both continue to rise.

Elsewhere (Tang 2009b) I argue that international politics has always been an evolutionary system, and it had firmly evolved from an offensive realism world before WWI to a defensive realism world after WWII. As such, we are living in a defensive realism world today, and defensive real-ism should be an appropriate guide for states in managing their security. If this is the case, then a coherent statement on defensive realism contributes to not only the growth of scientific knowledge but also the practical goal of managing security today.

The Task of This Book

For the debates among the major grand theories of international politics to be more productive, we need every grand theory to advance a coherent, if not systematic, statement about itself. A coherent statement of a grand the-ory must explicitly spell out: (1) the core assumptions of the grand theory; (2) the linchpin concept(s) of the grand theory; (3) how the grand theory stands differently from other grand theories in the two core issue areas of international politics—that is, war and peace, conflict and cooperation; and (4) how the grand theory expects states to behave under some gener-alized circumstances—an “operational code” for states’ strategic behav-iors.3 Without a coherent statement from each grand theory, it becomes extremely difficult to compare the validity of different grand theories, and a grand theory and its critics will often talk past each other.

introduction / 3

Among the three major grand theories, however, it seems that only John Mearsheimer (1994–1995; 2001) has taken on the task of formulating a coherent statement for offensive realism.4 In contrast, despite many faith-ful followers, defensive realism has nothing that comes close to a coherent, not to mention a systematic, statement, although Charles Glaser (1994), Robert Jervis (1978; 1999), and Jeffrey Taliaferro (2000–2001) have all moved us toward such a goal significantly.

Existing works of defensive realism fall into two general categories. The first type underscores the differences versus other theories in a piecemeal way, thus leaving many critical issues unaddressed and unresolved. The second type develops and tests specific mid-level theories. While defensive realists’ persistent and vigorous defense of defensive realism has greatly buttressed defensive realism’s claim as a grand theory of international pol-itics, such fragmented expositions are inadequate if defensive realism is to be respected as a grand theory of IR and a fitting theory for guiding states’ security strategies. We need a coherent statement of defensive realism. And because defensive realism has a close sibling in offensive realism, such a coherent statement necessarily includes a rigorous demarcation of offensive realism and defensive realism.

Lacking a coherent statement of defensive realism and a rigorous demar-cation of offensive realism and defensive realism has some unpleasant con-sequences, theoretically and policy wise.

First, efforts to develop mid-level theories have often gone astray. For instance, Taliaferro (2004) tried to combine defensive realism and pros-pect theory to explain why great powers got involved in peripheral areas and tended to stick with a failing project. Taliaferro, however, failed to notice that Germany before 1905 and Japan in the 1930s—the subjects of two major cases in his book—were both hardcore offensive realist states and that defensive realism does not apply to them. Moreover, if prospect theory captures a universal human psychological trait of loss aversion, then it should apply to both offensive realist states and defensive realist states (I define the two types of states in chapter 1). If so, then combining offensive realism with prospect theory would have explained the two cases much better than Taliaferro’s thesis has managed.

Second, some obvious mistakes have been committed in theoretical dis-cussion. For instance, Jeffrey Taliaferro identifies that domestic politics can limit the efficiency of a state’s response to the external environment as an auxiliary assumption unique to defensive realism, thus implying that as long as a realism theory admits a role for domestic politics in shaping state’s security behaviors, that theory is a defensive realism theory (Taliaferro 2000–2001, 131; see also Legro and Moravcsik 1999; Rendall 2006, 524–545). Likewise, when trying to explain Britain’s policy toward the


German threat before WWI, Steven Lobell asserted that offensive realism “black boxes the state” and “disregards how a state’s external security strat-egy is driven by international political competition” (Lobell 2002–2003, 166). As becomes clear in chapter 1, such an interpretation of offensive realism is incorrect: neoclassical theories within the offensive realism par-adigm do study domestic politics (e.g., Snyder 1991; Zakaria 1998). Both Taliaferro and Lobell mistakenly equated offensive realism with structural realism, and neoclassical realism with defensive realism.

Third, lacking a coherent statement of defensive realism and a rigorous demarcation of offensive realism and defensive realism even renders some criticism of realism wanting. For example, failing to understand the fun-damental difference between offensive realism and defensive realism (and realisms at different levels), Jeffrey Legro and Andrew Moravcsik (1999) asserted that any (realism?) theory that does not predict states to pursue expansions in earnest and almost reckless fashion does not quality for a realism theory. As such, they ended up in expelling much of the defensive realism literature from realism and advanced a position strikingly simi-lar to offensive realism, while claiming that their formulation of realism can accommodate both offensive realism and defensive realism.5 Echoing Legro and Moravcsik, Matthew Rendall contended that offensive realism predicts states to pursue expansion against great odds and that any trace of domestic politics is incompatible with offensive realism (Randall 2006, 526). Hence, for him, U.S. expansion on the North American continent was consistent with defensive realism because the expansion was expansion into a vacuum (525). Yet, U.S. expansion on the North American continent was exactly what offensive realism should predict: the United States in its continental expansion was the “poster child for offensive realism” (Layne 2002–2003; see also Zakaria 1998; Mearsheimer 2001; Elman 2004).

Last and most importantly, lacking a coherent statement of defensive realism and a rigorous demarcation of offensive realism and defensive real-ism makes it difficult to draw practical implications for states’ grand strat-egies even if statesmen want to heed advices from IR theorists.

We therefore need a coherent statement of defensive realism. This book seeks to fill this void.

The Approach of the Book

Nowadays, a typical IR book generally has the following structure. The first chapter introduces the puzzle to be explained, offers a critique of exist-ing explanations, and promises to offer something better. The second chap-ter lays out the author’s theoretical framework (i.e., a mid-level theory). The next few chapters will then be devoted to test the author’s theoretical

introduction / 5

framework against other alternative explanations with empirical evidences. The final chapter summarizes the findings and then discusses the theoret-ical and policy implications of the findings. Such a structure is adequate and even necessary if the goal is to advance a mid-level theory. But when the task is to formulate a coherent statement of a grand theory, such an approach is inadequate and thus inappropriate. Quite plainly, because a grand theory—by definition—must accommodate many mid-level theo-ries, one book simply will not be enough if one is to test all the mid-level theories within a grand theory.

Because the main purpose of this book is to advance a coherent statement of defensive realism, this book will be an almost purely a theoretical exercise. I shall make no serious attempt to test and support any of the mid-level theories espoused in the book with empirical evidences. Such an approach actually serves important purposes for theory-building, for three reasons.

First, a coherent statement of a grand theory advances a clear demarca-tion of the grand theory versus other competing grand theories, facilitating the debate among different grand theories. Second, a coherent statement of a grand theory involves making some previously implicit and unartic-ulated arguments and assumptions within the literature more explicit. Doing so facilitates further theory-building because many areas of confu-sion in international politics often result from implicit and unarticulated arguments and assumptions (Powell 1994). Third, a coherent statement of a grand theory assesses major developments within the grand theory. Such a job, though dirty it may be, is necessary because it can redirect theoret-ical developments.

The Structure of the Book

The book is divided into three major parts, “Preparing the Ground,” “Strategic Choices,” and “Implications and Conclusions.”

The first part, “Preparing the Ground,” consists of chapters 1, 2, and 3. These three chapters are an extended critical reading into the existing lit-erature that examines some of the core issue domains (e.g., the relationship between power and security) and concepts (e.g., the security dilemma) in defensive realism and realism in general. Together, these three chapters pave the way toward a coherent statement of defensive realism as a theory of security strategy for our time.

Chapter 1, “Clearing the Theoretical Underbrush,” lays down part of the conceptual foundation. Most prominently, I expose several false demarcation lines and reaffirm several genuine demarcation lines between offensive realism and defensive realism. I also advocate a more rigorous dichotomy for labeling two archetypes of states.


Chapter 2, “The Security Dilemma: A Conceptual Analysis,” critically revisits the linchpin concept of defensive realism, the security dilemma. This chapter defines the security dilemma more rigorously and clarifies some critical areas of confusion regarding the concept.

Building on chapter 2, chapter 3, “The Security Dilemma and War Revisited,” addresses the critical puzzle of security dilemma and war. This chapter unequivocally argues that the security dilemma cannot possibly have been a major cause of war, from both an empirical perspective and an evolutionary perspective. This chapter further emphasizes that exactly because the security dilemma has not been a major cause of war, the secu-rity dilemma is a powerful analytical tool for understanding and manag-ing international politics today. This seemingly contradictory notion holds because international politics has firmly evolved from an offensive realism world in which the security dilemma had been rare to a defensive realism world in which the security dilemma is now prevalent (Tang 2009b).

The second part of the book, “Strategic Choices,” consists of chapters 4 and 5. Building on the chapters in part I, this part advances toward a coherent statement of defensive realism as a theory of security strategy. Chapter 4 lays down a general operational code of defensive realism, or a coherent statement about what a defensive realist state should do under some generalized circumstances. Chapter 5 advances a coherent statement on reassurance as a defensive realism theory of cooperation-building. This chapter sheds new light on some of the key issues in reassurance: the rela-tionship between signaling of one’s own benign intentions and reading others’ intentions, the credibility of reassurance signals, and various forms of reassurance gesture.

Chapter 6 recategorizes realism theories more rigorously, focusing on those theories that have been inadequately or incorrectly understood. Among other things, this chapter makes it clear why some seemingly neo-liberalism theories are actually defensive realism theories and why Arnold Wolfers should be considered to be the first quintessential defensive realist.

A concluding chapter draws out some broader policy implications for states’ security policies from a defensive realism perspective.

Part I

Preparing the Ground


Chapter One

Clearing the Theoretical Underbrush

To advance a coherent statement of defensive realism, we need to first define what defensive realism is. This is no easy task, primarily because there have been so many realisms out there, even if we merely restrict our scope to IR. These include structural realism, neorealism (Waltz 1979), modi-fied realism (Krasner 1982a), classical realism, modern realism (Frankel 1995–1996), neoclassical realism (Rose 1998), post-classical realism (Brooks 1997), contingent realism (Glaser 1994–1995), cultural realism (Johnston 1995), offensive/aggressive realism, defensive realism (Snyder 1991), Bayesian realism (Kydd 2005), general realism, specific realism (Rosecrance 2002), hardcore realism, soft realism (Larson 1997), moti-vational realism (Kydd 1997b), and human nature realism (Mearsheimer 2001, 19–22). No wonder Glenn Snyder simply gave up counting (Snyder 2002, 149–150).

The proliferation of realisms can be extremely confusing even to stu-dents of IR, not to mention statesmen, policy analysts, and the general public. The piling up of realisms is so disheartening that two critics of realism, in a now “infamous” article, asked rhetorically, “Is anybody still a realist?” (Legro and Moravcsik 1999).1 To advance a coherent statement of defensive realism, we have to first sort out the numerous “realisms.”

In the first section of this chapter I provide a foundation for sorting out the various forms of realism. In the second and third sections I underscore the false and real demarcation lines between offensive and defensive real-isms. In the fourth section I explicate the serious defects in the existing labels for the two types of states—offensive realist states and defensive realist states. Last, I underscore the overarching fundamental difference between the two realisms and advocate more rigorous labeling of the two types of states and briefly discuss the implications.


Sorting out Realisms

To sort out the various realisms, I employ the strategy of “trickling-down”: I start with realisms at the broadest level and then gradually add more spe-cific assumptions to arrive at realisms at a narrower level.

From Political Realism to Realism in International Politics

Political realism, or realism toward politics, is not a theory or even a regular paradigm, it is a super-paradigm. Political realism is essentially a “philosophi-cal position [toward politics]” (Gilpin 1996, 6; see also Frankel 1996b, xiii).

Political realism starts with four bedrock assumptions: (1) The nature of politics is fundamentally conflictual; (2) Actors in politics are strategic;2 (3) Power is the primary yardstick and thus a fundamental feature of politics—that is, the struggle for power is at the heart of politics; (4) Material forces (as power) largely determine outcomes in politics (Legro and Moravcsik 1999, 16–18; Feaver et al. 2000, 181).3 These four assumptions as a whole underpin political realism as a coherent theoretical paradigm.4 The first three assump-tions dictate the nature of politics, the nature of actors in politics, and the general proxy goals of politics (i.e., power). The fourth assumption restricts how far states can get away in pursuing bad material or nonmaterial goals.5

Within political realism, we can further differentiate political real-isms in two very different social domains: domestic politics and interna-tional politics. The difference between the two domains, as Waltz (1979) has emphasized most forcefully, is that domestic politics is hierarchical whereas international politics is anarchical. Thus, for political realism in international politics, one additional assumption will be added to political realism: international politics is anarchical.6

Finally, to differentiate from liberalism that takes individuals and pri-vate groups to be the basic units of politics (Moravcsik 2003, 161–163), political realism in international politics adds still another critical assump-tion: collective units (i.e., states or other independent political entities) are the principal actors of international politics (Gilpin 1996, 7; Schweller and Wohlforth 2000, 69–70).

Altogether, political realism in international politics, and simply “real-ism” hereafter, starts with six bedrock assumptions (see table 1.1 for a sum-mary). With this definition in hand, we can move to sort out the various forms of realism in IR.

Theories at Different Levels of Analysis

Waltz (1959) did a great service to the study of international politics by making it explicit that there are three basic levels of analysis, or three

clearing the theoretical underbrush / 11

“images,” in international politics: the individual level (the first image), the state level (the second image), and the structural level (the third image). His three levels of analysis provide us with the first dimension to address the relationships among the various forms of realism.

Taliaferro (2000–2001, esp. 131–135) made the first attempt to advance a systematic categorization of realisms, explicitly along “the two crosscut-ting divisions with realism” (i.e., levels of analysis and the offensive real-ism versus defensive realism divide). More recently, Liu and Zhang (2006) improved on Taliaferro’s scheme but retained most of his analysis. Despite making many valid points, however, both Taliaferro’s and Liu and Zhang’s understandings of different strains of realism are incomplete and often incorrect.

Taliaferro misunderstands the nature of theories centered on differ-ent levels of analysis. Whereas he insists that theories at different levels of

Table 1.1 Political realism and realisms

Political Realism: Four Core Assumptions1. The nature of politics is fundamentally conflictual.2. Actors in politics are strategic.3. Power is the fundamental feature of politics.4. Political outcomes are determined primarily by material forces.

Political Realism in International Politics: Two additional assumptions1. International politics is anarchical.2. Collective units (e.g., states) are the principal actors of international politics.

Political Realism in International Politics: Various Versions1

Offensive RealismAdditional assumption: States are inherently aggressive (whether by nature or because of anarchy).

Defensive RealismAdditional assumption: States are not inherently aggressive.2

Structural realism Structural offensive realism

Structural defensive realism

Neoclassical Realism Neoclassical offensive realism

Neoclassical defensive realism

Individual level realism: human nature realism

Human nature offensive realism

Human nature defensive realism

Individual level realism: the social psychological school

Social psychological offensive realism

Social psychological defensive realism

Notes : 1 Chapter 6 and table 6.1 provide examples of different theories. 2 As such, there is no real uncertainty with states’ intentions for offensive realism (Tang 2008a).


analysis mostly distinguish themselves by seeking to explain outcomes at different levels (Taliaferro 2000–2001, 132), theories at different levels of analysis in fact distinguish themselves mostly by explaining outcomes with factors at different levels (Liu and Zhang 2006, 113–117; see also Glaser 1994–1995, 50n1).7

Specifically, Taliaferro claims that structural realism is the type of real-ism that primarily seeks to explain outcomes of international politics at the structural level (e.g., war and peace, stability of international system). However, structural realism does seek to explain some patterns of state behavior. For instance, in addition to the stability of bipolarity, Waltz also seeks to explain the recurrence of balancing behavior, which can be understood as both a structural-level outcome and a state-level outcome because balancing requires states to take purposeful actions (Waltz 1979, Chapter 7),8 and he emphasizes that balancing behavior merely requires anarchy and the instinct to survive among units to operate. Likewise, Mearsheimer (2001) singles out anarchy as the ultimate driving force behind great powers’ quest for hegemony, which leads to the tragedy of great power politics. Similarly, Copeland’s theory of dynamic differentials and war relies on three structural variables—polarity, the size of the dif-ferentials of power, and the trend of those differentiations—to explain preventive wars among great powers (Copeland 2000, 15, 29). What is common to these three theories is that they all rely (almost) exclusively on structural variables as independent variables.9

Neoclassical realism denotes some of the more recent works that seek to develop a theory of foreign policy (Rose 1998; Lobell, Ripsman, and Taliaferro 2009). Like classical realism, neoclassical realism seeks to explain certain state behaviors. Unlike classical realism, which usually relies exclu-sively on state-level variables to explain state behavior, however, neoclas-sical realism recognizes the impact of structural forces on state behavior. At the same time, neoclassical realism is also dissatisfied with structural realism’s inability (and sometimes, unwillingness) to explain states’ behav-ior due to its exclusive reliance on structural variables. Neoclassical realism thus seeks to bring structural variables and state-level variables together in order to develop a more coherent theory of state behavior.10

In additional to structural and neoclassical realism, there has always been a strain of realism that tends to rely on—though not exclusively—factors at the individual level for explaining state behavior, although these theories do not seem to share a coherent “core.”11 For the most part, this strain of realism relies on various (fundamental) drivers of human behavior: an insatiable lust for power or domination (e.g., Machiavelli, Morgenthau), a fear of violent death (e.g., Hobbes), and honor or glory (e.g., Hobbes, Machiavelli).12 More recently, Thayer (2004) added a sociobiological twist


to this tradition, by claiming that human beings fight because conflicts increase personal and group fitness.13 Taliaferro’s integration of prospect theory with defensive realism also belongs to this category of theories (Taliaferro 2004).

Other than (classical) human nature realism, there also exist realism theories at the individual level that do not rely on a sweeping general-ization about human nature but on more specific psychological traits of human cognition. This strain of realism can be perhaps dated back to Alexander George’s work on “operational code” (George 1969), but it has flourished largely due to the social psychological school toward IR, pio-neered most prominently by Jervis (Jervis 1970, 1976; Janis 1977; Khong 1992; Larson 1997). More recently, this strain of realism has tended to produce more specialized work, most prominently represented by works on reputation (credibility) and war (McMahon 1990; Mercer 1997; Press 2005; Tang 2005). In addition, Byman and Pollack (2001) made a system-atic effort to formulate a more rigorous framework for understanding the role of individuals in international politics, and Samuels produced a vol-ume that focuses exclusively on the effect of individuals on states’ behavior (Samuels 2003).

The major difference between these realisms at different levels of analysis is that they place different weight on structural, state-level, and individual-level factors for explaining states’ behavior and international outcomes. Structural realism relies almost exclusively on structural factors, human nature realism almost exclusively on individual-level variables, and neoclassical realism on a mixture of structural variables and state-level variables, and sometimes even individual-level variables.

Finally, a caveat is in order. Emphasizing these three levels of analysis does not mean that they are mutually exclusive. Since the mid-1980s, more and more scholars have increasingly recognized that an adequate under-standing of international politics, be it international outcome or state behavior, requires bringing factors at different levels of analysis (and dif-ferent approaches) together. Indeed, Waltz himself has repeatedly empha-sized that “some combination of our three images, rather than any one of them, may be required for an accurate understanding of international relations” (Waltz 1959, 14, see also 160, 225, 238).14

The Fundamental Bifurcation: Offensive versus Defensive Realism

Realism and other major theories of international politics owe another enor-mous debt to Waltz: the coming of Waltz’s neorealism reinvigorated the the-oretical debates in international politics because Waltz’s bold and sweeping claim provided the perfect target at which everybody can take a whack.


It is within this new wave of debate that a fundamental bifurcation within the realism camp, the offensive realism and defensive realism divide, began to firmly emerge.15 Unlike realisms at different levels of analysis that are complementary to each other (Taliaferro 2000–2001, 132; Liu and Zhang 2006, 113–117), offensive realism and defensive realism are two competing theoretical paradigms. These two strains of realism, even if they focus on the same level of analysis, reach fundamentally divergent interpretations of international politics.

Offensive versus Defensive Realism: False Fault Lines

Because offensive realism and defensive realism seem to subscribe to the same set of bedrock realist assumptions and the two realisms have long used the same vocabulary even though the same word or phrase may mean quite different things for them, much conceptual and logical confusion reigns. As a result, the differences between the two realisms have been either inad-equately recognized or outright misconstrued, despite numerous earlier attempts (e.g., Brooks 1997; Frankel 1996; Glaser 1994–1995; Jervis 1999; Kydd 2005, Chapter 1; Labs 1997, 7–17; Layne 2006, Chapter 1; Liu and Zhang 2006; Lobell 2002–2003, 165–176; Lynn-Jones 1998; Schweller 1996; Snyder 1991; Taliaferro 2000–2001; 2004, 11–18).

To get to the real differences between the two realisms, we first need to be clear what does not constitute the critical differences between the two realisms. Here, two previous attempts by Brooks (1997) and Taliaferro (2000–2001) to rigorously differentiate the two realisms should be noted. These two authors reached similar conclusions: the fundamental differ-ences between the two realisms are caused by some auxiliary assumptions. For them, defensive realism admits those auxiliary assumptions while offensive realism does not, either for the sake of parsimony or because offensive realism cannot survive with those auxiliary assumptions (Brooks 1997, 470; Taliaferro 2000–2001, 131, 134–143).

Brooks (1997)

The three auxiliary assumptions identified by Brooks are: (1) possibility versus probability; (2) short term versus long term; and (3) military secu-rity versus economic capacity. Brooks (1997, 450) further noted that these three assumptions can be subsumed under the problem of possibility ver-sus probability.

Brooks’s formulation is under-specified, thus too simplistic, and it can be easily refuted. Brooks did not specify that offensive realism is a possibilistic theory only when it comes to other states’ intentions but probabilistic when


it comes to other states’ capabilities, interest, and resolve. In other words, offensive realism only assumes the worst over others’ intentions, but not over others’ capabilities, interest, and resolves.16 Moreover, because offen-sive realism also assumes states as unitary rational actors (Mearsheimer 1994, 10; 2001, 21–22), and acting “rationally” is generally understood to be acting according to probability, with possibility being an extreme expression of probability, Brooks’ under-specified formulation is inconsis-tent with realism’s bedrock assumptions. Finally, Brooks’s formulation sug-gests that only defensive realist states face tradeoffs between long-term and short-term, between military security and economic well-being, between competition and cooperation. This formulation is incorrect because offen-sive realist states also face those tradeoffs.17

Taliaferro (2000–2001)

The four auxiliary assumptions listed by Taliaferro are: (1) the security dilemma is an intractable feature of anarchy; (2) “structural modifiers” (e.g., the offense-defense balance, geography) influence the severity of the security dilemma between particular states; (3) material power drives states’ foreign policies through the medium of leaders’ calculations and per-ceptions; (4) domestic politics can limit the efficiency of a state’s response to the external environment. These assumptions do not constitute valid demarcation lines either, because theories within the offensive realism camp have not consistently excluded these auxiliary assumptions.18

The Security DilemmaBecause of the centrality of the security dilemma in defensive realism, it was perhaps natural for Taliaferro to identify whether the security dilemma is an intractable feature of anarchy as the first auxiliary assumption that differentiates offensive realism from defensive realism. Offensive realism denies the existence of the security dilemma, whereas defensive realism holds that the security dilemma is an intractable feature of international politics (Taliaferro 2000–2001, 131, 136). Taliaferro identified Schweller’s 1996 article “Neorealism’s Status Quo Bias: What Security Dilemma?” as an important work of offensive realism because Schweller explic-itly contended that the security dilemma is not an intractable feature of international politics but instead exists only under certain circumstances (Schweller 1996, 117; Taliaferro 2000–2001, 144–147).19

Taliaferro’s stand seems to be neat and straightforward, because some proponents of offensive realism tend to neglect the security dilemma. For instance, Labs (1997) did not mention security dilemma in his theory of the expansion of war aims. Likewise, despite citing Herz and Jervis’s work


on the security dilemma, Zakaria (1998) was agnostic about whether the security dilemma exists and its implications for states’ behavior (the secu-rity dilemma was not indexed in his book).

Contrary to Taliaferro, however, some proponents of offensive realism do admit the security dilemma, often emphatically.20 Mearsheimer, the quintessential offensive realist, has not only admitted that the security dilemma is an intractable and prevalent feature of international politics but has used the security dilemma to bolster his offensive realism argu-ments (Mearsheimer 2001, 35–36). The security dilemma also features prominently in Copeland’s theory of dynamic differentials and preven-tive war (Copeland 2000, 13, 24–25, 36–37). Yet Copeland’s theory is an offensive realism theory, although Taliaferro incorrectly categorized it as a defensive realism theory. Finally, in an indirect critique of defensive realism and defense of offensive realism, Keir Leiber argues that “offensive realism . . . sees the security dilemma as an immutable feature of interna-tional politics so long as states operate in anarchy,” but then also argues that “offensive realists reject the security dilemma as a model of interna-tional politics (Lieber 2005, 8, 156).”

Moreover, a quintessential defensive realist, Christopher Layne (2006, 15) explicitly called the security dilemma “a misnomer” and gen-erally dismissed it. Yet, Layne is undoubtedly a defensive realist because he explicitly argues that seeking hegemony, a strategy supported by all offensive realists (e.g., Mearsheimer 2001; Zakaria 1998), will be counter-productive because it causes counterbalancing or “geopolitical backlash” (Layne 2006, 5). Layne further calls for the United States to be an offshore balancer, a strategy that preaches self-restraint and moderation.21

Acknowledging that the security dilemma is, or is not, an intractable fea-ture of international politics cannot neatly delineate offensive realism and defensive realism. As I will show in chapter 4, the real difference between the two realisms on the security dilemma lies in the fundamentally differ-ent implications that they draw from the existence of the security dilemma and the diametrically opposite strategies that they recommend on what to do about it. Offensive realism argues that states should try to escape from the security dilemma through offensive actions (e.g., annihilating the other state), whereas defensive realism argues that states should try to alleviate the security dilemma through cooperative actions (e.g., reassur-ance and cooperation).

Structural ModifiersBy listing the notion that “structural modifiers” (e.g., offense-defense bal-ance, geography, and access to raw materials) influence the severity of the security dilemma between particular states as another auxiliary assumption


that is unique to defensive realism, Taliaferro implies that offensive realism simply ignores the effect of “structural modifier.”22

Yet, by emphasizing the stopping power of a large body of water against war and conflict, Mearsheimer implicitly admitted that at least one struc-tural modifier (i.e., geography) does influence the possibility of war and conflict (Mearsheimer 2001, 41). More evidently, traditional geopolitics, a prominent variant of offensive realism, not only emphasizes the impact of geography on state behavior but comes close to being “geographical deter-minism” (see chapter 6 this volume).

Moreover, both Mearsheimer (2001) and Copeland (2000) explicitly acknowledge not only the existence of the security dilemma but also the impact of nuclear weapons on state behavior. For Mearsheimer, nuclear weapons make global hegemony seem impossible. Likewise, Copeland emphasized that nuclear weapons mean that preventive war by a reigning hegemon is no longer a viable option, and a far more plausible scenario will be that a reigning hegemon initiates crises in order to forestall the growth of a rising power. Because nuclear weapons are generally understood as the ultimate weapons that have brought defense dominance to international politics (Jervis 1989), Mearsheimer and Copeland obviously acknowl-edged that the offense-defense balance influences the security dilemma and state behavior by admitting that nuclear weapons make certain things difficult.

Leaders’ Calculations and PerceptionsThe third auxiliary assumption that Taliaferro identified as being unique to defensive realism is that material power drives states’ foreign policies through the medium of leaders’ calculations and perceptions. Taliaferro implies that offensive realism sees no role for leaders’ calculations and per-ceptions and believes that there is a straightforward correlation between material power and state behavior.23

Yet, the state-centric theory developed by Zakaria (1998), which Taliaferro correctly classifies as an offensive realism theory, explicitly centers on leaders’ calculations and perceptions (about the ability of the state to extract resources or mobilize national power). Similarly, Eric Labs’ theory of expanding war aims also depends on leaders’ calculations and perceptions: it would be close to impossible to understand why Prussia decided to annex Alsace-Lorraine without taking into account Bismarck’s bitterness toward France (Labs 1997, 28–34). Finally, even Copeland’s theory of dynamic differentials and war, which is mostly a structural offensive realism theory, implicitly integrates leaders’ perceptions (about whether relative decline is deep and inevitable) into its theoretical frame-work (Copeland 2000).


Domestic PoliticsTaliaferro argues that as long as a realism theory admits a role for domestic politics in shaping state’s security behaviors, that theory is a defensive real-ism theory. This take on the role of domestic politics in offensive realism and defensive realism is very similar to the notion that offensive realism does not take leaders’ calculation and perceptions into account, thus again mistaken.24

To begin with, to identify the notion that domestic politics can limit the efficiency of a state’s response to the external environment as an auxil-iary assumption of defensive realism is to deny that there can be structural theories within the defensive realism camp. Herz (1951), Jervis (1978), and Glaser (1992, 1994–1995) have all developed structural theories within the defensive realism camp, and their theories do not pay much attention to domestic politics, if at all.25 More prominently, Waltz’s balancing-of-power theory, which Taliaferro classified as a defensive realism theory, too has no trace of domestic politics.

Furthermore, while structural theories of offensive realism do not pay much attention to the role of domestic politics in shaping state behavior and international outcomes (Copeland 2000; Mearsheimer 2001), there are theories within the offensive realism camp that emphasize the role of domestic politics for understanding state behavior and international out-comes. Again, Zakaria’s state-centric realism, an offensive realism theory, explicitly argues that, for much of the later nineteenth century, the United States wanted to expand but was held back by a relatively weak state that was unable to extract and mobilize national power (Zakaria 1998).26 Likewise, Jack Snyder’s logrolling theory of imperialism, another quintes-sential neoclassical theory of imperialism (or offensive realism) and impe-rial overextension, pays tremendous attention to domestic politics (Snyder 1991; see also chapter 6 this volume). Finally, Schweller (2006), whom Taliaferro classifies as a quintessential offensive realist (because Schweller argued that the security dilemma is conditional), has been one of the most vocal defenders of bringing in unit-level variables.

Whether a theory takes domestic politics seriously is a demarcation line between structural theories and state-level theories, not a demarcation line between offensive realism and defensive realism.27

Seeking (Relative) Power versus Seeking Security

I now examine a false fault line that is perhaps the most widely adopted in the realism literature, the notion that defensive realism believes that states seek security whereas offensive realism believes states seek power.

According to realism, every state is insecure, or at least feels inse-cure, under anarchy. In other words, all states are driven by fear. Thus,


in a nutshell, all states seek security. Because both offensive realism and defensive realism subscribe to the core assumption of realism that power is the fundamental feature of (international) politics and because politi-cal outcomes (e.g., states’ security) are primarily determined by material power, both offensive realism and defensive realism believe that seek-ing relative power is an important means of external self-help under anarchy.

As a result, offensive realists can still claim that their theory starts with the assumption that states seek security, and it just happens that their the-ory leads to the conclusion that maximizing relative power is the best and the only sure means toward security (Labs 1997, 4–5, 11; Mearsheimer 2001, 29, 32–36).

Thus, on seeking security or power as goals or means toward goals, the difference between offensive realism and defensive realism is not that offensive realism believes that states seek relative power whereas defensive realism believes that states seek only security but not relative power, as many have believed (e.g., Glaser 1995, 74–75; Legro and Moravcsik 1999, 16; Taliaferro 2001, 128–129). The two realisms contend that states seek both security and relative power (either as a goal in itself or as a means toward security). Because power remains an important pillar for security under anarchy and because power and security interact (Jervis 1988, 336; Layne 2006, 17), “there is no possibility of drawing a sharp line between the will-to-live and the will-to-power (Niebuhr 1960 [1932], 42).” The dif-ference between offensive realism and defensive realism thus cannot be captured by the notion that the former demands states to seek only relative power, whereas the latter demands states to seek only security.28

Offensive versus Defensive Realism: Real Fault Lines

In this chapter and the chapters that follow, I will show that the funda-mental differences between the two realisms lie mostly in their different preferences over strategies, or over what exactly constitute the viable means of self-help under anarchy.29 The two realisms’ differences over strategies are in turn underpinned by their differences on how to cope with the uncertainty over others’ intentions and the fear derived from it: Offensive realism holds that states must assume the worst over other states’ inten-tions whereas defensive realism (and other non-offensive realism theories) believes that states must not rashly assume the worst over other states’ intentions (Tang 2008a; see also Rathbun 2007).

Offensive realism and defensive realism agree that international poli-tics, without a central authority, is a realm of self-help. There is no 911 to call when a state’s security is threatened under anarchy, and each state has to take care of itself (Mearsheimer 2001, 32–33; Waltz 1979, 91, 111).


Waltz (1979, 105–107, 111–114, 118) divided the means of self-help into two general categories: internal efforts and external efforts. Regarding internal means of self-help, both offensive realism and defensive realism agree that internal balancing through economic growth is a means of inter-nal self-help (Mearsheimer 2001: 33–34; Waltz 1979, 118).30 The two real-isms diverge greatly on means of external self-help, but their differences on what constitute viable means of external self-help are far more subtle and complex than many have conceived. Moreover, because proponents of the two realisms have often advanced inconsistent and incoherent arguments, those differences have not been properly understood.

Common Interest and Cooperation

The first real fault line between the two realisms is common interest and cooperation.31 Offensive realism argues that cooperation among states under anarchy is inherently difficult and risky due to the concern for rela-tive gains, the temptation to cheat, the high cost of being cheated, and the difficulty of detecting cheating (Mearsheimer 2001, 51–53; see also Grieco 1988; Waltz 1979, 105–106; 1986, 336). As a result, offensive realism does not consider cooperation, other than temporary alliance when facing a common threat, as a viable means of external self-help.

In contrast, defensive realism argues that cooperation is a viable means of external self-help and that it is possible under a wide range of circum-stances (Glaser 1994–1995; Jervis 1999). Defensive realism contends that, at the very least, there is no compelling logic embedded in realism’s bed-rock assumptions to exclude cooperation as a viable means of external self-help (Glaser 1994–1995, 51, 58–60).

Although correct, this formulation does not really get to the bottom of the question, because the two realisms’ difference on cooperation arises from a more fundamental difference that has been generally unrecognized in the literature: common interest.

Specifically, offensive realism sees that states are inherently aggressive— either by nature or because of anarchy—and their aggressiveness is limited only by their capabilities (Mearsheimer 2001, 2–3, 21, 32–36). As such, there is very little common interest among states, other than temporary common interests when facing a common threat. Consequently, states should see little rationale for cooperation.

In contrast, defensive realism rejects the notion that states are inher-ently aggressive by nature or that anarchy compels states to be inherently aggressive. As such, two states do not have to intentionally threaten each other. When this is the case, the two states have at least the common inter-est in avoiding costly competition (e.g., arms race, wars) between them


and cooperation between them becomes possible and profitable for their security. Under such a circumstance, cooperation is a viable and critical means of external self-help (Glaser 1994–1995; Jervis 1999).

Hence, a more fundamental assumption—whether there is common interest under anarchy—has underpinned offensive realism’s and defensive realism’s different stands on cooperation. Whereas offensive realism sees little or no common interest among states, defensive realism does sees some and perhaps even significant amount of common interest among states. In this sense, it is redundant for offensive realism to argue that cooperation is inherently difficult due to states’ concern for relative gains, the tempta-tion to cheat, and the high cost of being cheated. In an offensive realism world, most strategic situations among states are, in game theoretical jar-gon, “deadlocks” in which cooperation is the worst or second-worst out-come for states (Downs, Rocke, and Siverson 1985, 121). As such, there is no good reason for states to cooperate or even to try to cooperate.32

Any Danger Associated with Maximizing Relative Power?

The second real fault line between offensive realism and defensive realism is whether there is any danger associated with maximizing relative power. Offensive realism does not see any danger associated with maximizing rel-ative power, whereas defensive realism does.

By eliminating cooperation as a viable means of self-help and by taking alliance as a temporary solution (Mearsheimer 2001, 33), offensive realism comes down to seeing maximizing relative power as the only sure external means of self-help (Gilpin 1981, 23–24; Labs 1997, 4–5, 11; Mearsheimer 2001, 2, 32–36; Zakaria 1998, 19–22).33 As such, offensive realism does not see any danger associated with (successfully) maximizing relative power. Indeed, offensive realism contends that the danger lies in not actively seek-ing and grabbing opportunities of maximizing relative power. Because a state can never be sure how much power is enough for its security, it must try to gain as much power as it can (Labs 1997; Mearsheimer 2001, 32–36).

On the contrary, defensive realism emphasizes that maximizing rel-ative power—even when it is (temporarily) successful—is not always a good means of external self-help (Glaser 1994–1995, 71–72; Waltz 1979, 118, 126–127; Wolfers 1952, 490, 496–498). For defensive realism, a relentless maximization of relative power can be counterproductive, for two reasons. First, states tend to balance against power or threat. Hence, a state that relentlessly seeks relative power will appear so threatening to other states that they will counterbalance it and eventually make its relentless quest for power self-defeating. Second, the existence of the security dilemma also makes a relentless maximization of relative power


self-defeating (see chapter 2 this volume). Defensive realism believes that these two dynamics often work together to make maximizing relative power ultimately self-defeating sometimes.34

Are Seeking Relative Power and Seeking Security Essentially Equivalent?

The third real fault line between offensive realism and defensive realism is whether one can treat maximizing relative power and seeking security as essentially equivalent. Offensive realism does so whereas defensive realism does not.

Within the offensive realism camp, the picture is relatively clear. Even though some proponents of offensive realism argue that states simply seek relative power or “influence” (Zakaria 1998, 29–30),35 other propo-nents argue that anarchy drives states to seek security and that the best way toward security is to maximize relative power (Labs 1997, 4–5, 11; Mearsheimer 1994, 9–13; 2001, 2, 29, 32–36), the difference between these two positions is rather minor. One uses security-seeking as the inter-mittent link between anarchy and seeking relative power and the other does not, yet both arrive at the same argument: The only sure way toward security under anarchy is to maximize relative power because all other external means (e.g., alliance and cooperation) are unreliable. Because offensive realism believes that the only sure way toward security under anarchy is to maximize relative power, maximizing relative power is essen-tially equivalent to seeking security for offensive realism.36

In contrast, because defensive realism recognizes cooperation as a viable means of external self-help, it denies that maximization of relative power and seeking security are essentially equivalent (Lynn-Jones 1998, 177n47; Wolfers 1952, 485n4).37

The situation within the defensive realism camp, however, is more con-fusing, mostly due to Waltz’s stand on the so-called absolute versus relative gains problem in international cooperation.38 On the one hand, Waltz explic-itly emphasized that maximizing relative power may or may not serve the end of security, thus implying that maximizing relative power cannot possi-bly be the only sure external means of self-help (Waltz 1979, 118, 126–127; 1986, 334). On the other hand, however, Waltz (1979, 105) also explicitly stressed that states will be concerned with relative gains when considering whether to cooperate or not and that states’ concern for relative gains from cooperation will make cooperation difficult, if not impossible.39

Yet, why would a state always care for relative gains and reject coop-eration as a means of external self-help? The answer must be: (1) because relative gains can be readily translated into relative power (and power is


fungible, hence the more the better) and it is relative power alone that matters for security; and (2) because maximizing relative power is the only sure means toward security. If maximizing relative power is not always conducive to security, then states should not always be concerned with relative gains from cooperation. Thus, when Waltz’s stand on the concern for relative gains in cooperation is pushed to its logical conclusion, it must be the notion that maximizing relative power is the only means toward security that ultimately makes cooperation difficult. This conclusion, of course, contradicts Waltz’s other stand that maximizing relative power is not always conducive to security. Hence, although Waltz is often consid-ered a defensive realist, his understanding on relative power, security, and relative gains has been inconsistent and often self-contradictory.

In contrast, many defensive realists have argued that while concern for relative gains (or power potential) may make cooperation difficult some-times, it does not always do so (Glaser 1994–1995, 70–76; Powell 1991).40 This is fully consistent with defensive realism’s stance on the relationship between relative power and security: maximizing relative power may or may not serve the ends of security.

Unsatisfactory Labels

Much confusion also results from the fact that offensive realism and defensive realism have used some misleading, if not deeply flawed, dichot-omies for labeling offensive realist states and defensive realist states (see table 1.2 for details). These dichotomies are not only heuristic tools; they carry implicit or explicit messages that are of fundamental importance for understanding the core logic of offensive realism and defensive realism. This section critically examines these dichotomies and advocates a more rigorous dichotomy for labeling these two (ideal) types of state.

An important caveat is in order. Labels capture ideal types. As made more explicitly in chapter 2, states can change from one type to another as a combinatorial result of (1) state-to-state interaction, especially driven by the spiral generated by a security dilemma and (2) domestic politics, espe-cially change of leadership.41 As such, a state may occupy a gray zone and a particular label does not have to always stick.42

Revisionist State versus Status Quo State

Among all the dichotomies for labeling the two types of states, the dichot-omy of revisionist state versus status quo state has been the most promi-nent. Past critiques of this dichotomy tended to focus on its two drawbacks. First, the dichotomy provides a reigning hegemon with moral superiority


in rhetoric (Carr 1939, 105). Second, the dichotomy is almost impossi-ble to operationalize ex ante (Chan 2004).43 Despite making valid points, however, these criticisms missed the most fundamental drawback of this dichotomy: to equate a status quo state with a defensive realist state is sim-ply misleading, if not flatly wrong.44

In IR, “status quo” is merely the current distribution of power or a state’s current position within the current distribution of power. In other words, the status quo is merely a denominator for the result derived from the interaction between a state’s behavior and other states’ behavior ex ante, and it tells us nothing about the making or the nature of the status quo

Table 1.2 Dichotomies for labeling the two types of states

Offensive realist state Defensive realist state Examples

Commonly used but misleading labels

Revisionist state Status quo state Jervis (1978); Organski and Kugler (1980); Wolfers (1962); Schweller (1996, 1998); Mearsheimer (2001)

Revolutionary state Status quo state Kissinger (1957);Gaddis (1981)

Predatory state (predators)

Status quo state Schweller (1996); Kydd (1997b); Wendt (1992)

Imperialist/expansionist state

Stats quo state Morgenthau (1948); Snyder (1985)

Unsatiated/dissatisfied state

Satiated/satisfied state Carr (1939); Schumann (1948)

Useful but unsatisfactory labels

Relative power-seeker or maximizer

Genuine security-seeker or maximizer

Buzan (1991); Snyder (2002); Tang (2004)

Greedy state Not-greedy state Glaser (1992)Greedy state Security-seeking state Kydd (1997a);

Schweller (1996, 1998)Absolute security-seeker

Relative security-seeker

Chace and Carr (1988); Hendrickson (2002); Wolfers (1952)

Maximal security-seeker

Minimal security-seeker

Zakaria (1998)

Untrustworthy/Expansionist states

Trustworthy/security-seeking states

Kydd (2005)

The more rigorous label

Offensive realist state Defensive realist state

Useful label for public consumption

Malignant state Benign state Montgomery (2006)


and the characteristics of a state’s past, present, and future behavior. If so, whether a status quo state is a defensive realist state depends on what kind of status quo the state seeks to maintain and what kind of means the state employs to maintain the status quo.

For instance, Great Britain before WWI has often been labeled as a status quo power (thus implying that it was a defensive realist state), partly because it was the reigning hegemon. Yet, Great Britain was an imperialist power. It was constantly expanding its “defense” perimeter and it could not avoid impinging on other states’ interests, whether other states’ inter-ests were expanding or not.45

Moreover, a state that is striving to maintain the spoils of its past aggres-sion and conquest is no different from a state that is actively expanding or conquering. The former is as much an offensive realist state as the latter, and not a defensive realist state as usually implied by the label “stats quo state.” Put rhetorically, would anyone call a state a status quo power if it has just conquered someone else’s territory and then declared itself as such a power (even if this metamorphosis is true), if status quo power is to mean a defensive realist state? After all, a state can only build an empire by defending the spoils of its past aggressions and conquests (thus defend-ing the status quo for the moment). Here, a French source from 1799 is illuminating: “French leaders sought no new conquests; they wanted only to preserve the earlier gains of the Revolution” (Quoted in Jervis 1978, 186; emphasis added).46

Hence, even though Britain was a status quo power (i.e., it merely wanted to maintain its imperial leadership) before WWI, the status quo that it wanted to maintain—its far-fledging empire—was the result of past aggressions and conquests, and maintaining its preferred status quo required other states to be insecure or even absolutely insecure (i.e., becom-ing and remaining colonies). Likewise, Imperial Japan before WWII was not a defensive realist state (or security-seeking state for Taliaferro 2000–2001) but rather an offensive realist (expansionist) state, even though it too wanted to preserve the status quo (Taliaferro 2001, 147). For Imperial Japan, keeping Korea and the part of China it had earlier occupied was the status quo, and in order to maintain the status quo, Japan eventually had to expand into the whole Southeast Asia for strategic resources and fight against Britain and the United States. As such, neither Britain before WWI nor Imperial Japan before WWII was a defensive realist state but a quintessential offensive realist state, although both wanted to preserve the “status quo” and thus were “status quo states.”

Finally, according to prospect theory (Levy 1997, 90; Taliaferro 2004, 30–31, 227), loss aversion with a “status quo bias” is a universal psycholog-ical trait.47 Hence, it is not surprising that realists since Thucydides have


long noted that states loathe losing their current (relative) positions. For instance, Carr (1939, 111) noted that “the most serious wars are fought in order to make one’s own country militarily stronger, or, more often, to pre-vent another from becoming militarily stronger.”48 Similarly, Waltz (1979, 126) asserted that “the first concern of a state is not to maximize power, but to maintain their position in the system.” And Grieco put it most bluntly: “the fundamental goal of states in any relationship is to prevent others from achieving advances in their relative capabilities” (Grieco 1988, 498; emphasis in original). If loss aversion with a status quo bias is a universal psychological trait, then it cannot possibly differentiate defensive realist states from offensive realist states. Indeed, offensive realist states may hate losing ground even more.49

In other words, while a revisionist state is far more likely to be an offen-sive realist state,50 a status quo state is also likely to be an offensive realist state (Glaser 1992, 501n4; Wolfers 1952, 484n3). As a result of all these drawbacks, the dichotomy between a revisionist state and a status quo state simply cannot capture the differences between an offensive realist state and a defensive realist state.

(Relative) Power Maximizer/Seeker versus Security Maximizer/Seeker

The dichotomy of power maximizer/seeker versus security maximizer/seeker has also been popular (e.g., Schweller 1996, 108, 114; 1998, 24–25; Snyder 2002, 151–155; Taliaferro 2000–2001, 128–129; Tang 2004, 27; Wendt 1999, 104–105). This dichotomy too is logically and operationally unsatisfactory.

This dichotomy is explicitly or implicitly based on the misperception that offensive realist states seek relative power whereas defensive realist states seek security. Yet, according to realism, every state under anar-chy seeks both (relative) power and security. Moreover, because power and security do interact, “there is no possibility of drawing a sharp line between the will-to-live and the will-to-power” (Niebuhr, 1960 [1932], 42). Consequently, it is difficult to operationalize this dichotomy of power seeker versus security seeker for labeling different states.

Simply put, while a defensive realist state may claim that it seeks security rather than relative power, an offensive realist state can claim that it too seeks security yet it just happens to believe that maximizing (relative) power is the best and the only means of external self-help toward security. In fact, this is exactly the position that Mearsheimer’s offensive realist states would take (Mearsheimer 2001, 29, 32–36; see also Labs 1997, 4–5, 11).

For instance, while we can all agree that expansions for glory and pride are not exactly “security-seeking,” one can certainly make the case that


Germany in the two world wars and Japan before WWII were at least partly security-seeking. Germany wanted to escape from the predicament of its geographical location (i.e., the danger of having to fight a two-front war), and Japan wanted to make itself less dependent on imported natural resources (i.e., it wanted autarky). Thus, their behaviors were at least partly driven by security-seeking motives (Snyder 1991, 306–307).51 Yet, because they sought security by intentionally threatening others, it is hardly defen-sible to label them as “security seekers,” if the label of “security seeker” is to mean defensive realist states. Because security-seeking motive also (partly) drives the behavior of offensive realist states, it cannot differentiate offen-sive realist states from defensive realist states.

Now, some defensive realists may counter that their definition of secu-rity-seeking is different from that of offensive realists, saying that their definition means seeking security with a defensive approach (i.e., not intentionally threatening others).

But this is not the way those defensive realists operationalize their defini-tion of security-seeking motive. For instance, Glaser (1992, 1997) contended that the security dilemma is compatible with a greedy state (i.e., an offensive realist state) because even a greedy state might be insecure, and thus their behavior can be driven by security-seeking motives. More recently, to make the case that the security dilemma has been a major cause of war in history, Taliaferro (2001, 147–149) characterized many wars initiated by offensive realist states as security-driven expansions, implying that these wars were largely driven by the security dilemma. Apparently, Glaser and Taliaferro believe that a war or an act of expansion is driven by a security dilemma as long as it is driven by insecurity (i.e., security-seeking was the motive).

Yet, because a genuine security dilemma requires a lack of malign inten-tions from both sides to operate (see chapter 2 this volume), by arguing that the security dilemma is compatible with greedy states and character-izing many wars initiated by offensive realist states as wars driven by the security dilemma, Glaser and Taliaferro essentially argue that security-seeking does not necessarily mean lack of malign intentions or a defensive approach toward security. For Glaser and Taliaferro, security-seeking is compatible with both malign intentions and benign intentions.

As such, the dichotomy of (relative) power-seeker versus security-seeker, as it is understood in the literature now, cannot possibly differentiate between offensive realist states and defensive realist states.

Absolute Security-Seeker versus Relative Security-Seeker

A slightly better and perhaps more useful dichotomy has been the dichot-omy of absolute security-seeker versus relative security-seeker (Chace and


Carr 1988; Hendrickson 2002).52 Implicit in this dichotomy is that an absolute security-seeker—a state that seeks a very high sense of security—is an offensive realist state. Because absolute security can only be achieved by eliminating all other states in the system, seeking absolute security means that a state abhors the fact that its security can only be relative (because it depends on others) and it seeks to escape from this relativity. In contrast, a relative security-seeker is a defensive realist state because such a state under-stands that its security can only be relative (because it depends on others) and tries to live with this relativity (Wolfers 1952, 488, 492–494).

This dichotomy, however, is still wanting. The first and most outstand-ing drawback of this dichotomy is that even an extreme offensive realist state will admit that its security can only be relative because it cannot elim-inate all other states in the system and become a universal empire (Gilpin 1981, 94; Mearsheimer 2001, 41, 101–104). Hence, an offensive realist state can still claim that it is not seeking absolute security (not because it does not want it, but because it deems the goal impossible), but rather merely trying to improve or maximize its security. After all, if security is what a state wants, all states certainly should want more, and there is no reason why states should not wish their security to be absolute, even if it is impossible.

Second, while early defensive realists such as Bernard Brodie and Arnold Wolfers had taken it as self-evident that security can only be a matter of degree (Brodie quoted in Baldwin 1997, 14; see also Wolfers 1952, 484), some modern realists deny that security can be relative. For instance, Barry Buzan asserted that “the word [i.e., security] itself implies an absolute condition—something is either secure or insecure—and does not lend itself to the idea of a graded spectrum.” Buzan further contended that if secu-rity is conceived as a matter of degree (i.e., relative), “then a complicated and objectively unanswerable question arises about how much security is enough” (Buzan 1991, 18, 330). While David Baldwin disagreed with Buzan and argued that security should be conceived as relative, Baldwin gave only a normative reason—“absolute security is unattainable”—and did not provide us with a clue about how to operationalize the dichotomy (Baldwin 1997, 15).

Finally, by differentiating objective security and subjective security, Wolfers (1952, 485–489) explicitly recognized that different individuals can have very different assessments of a state’s security. This duality of sub-jectivity and objectivity poses a fundamental problem for operationalizing the dichotomy of absolute security-seeker versus relative security-seeker. How can one know ex ante whether a state seeks absolute security or rela-tive security, when both offensive realist states and defensive realist states seek security (Zakaria 1998, 26–27)?


Toward a More Rigorous Dichotomy

The preceding discussion reveals a common element between the false fault lines that try to demarcate offensive realism and defensive realism and the unsatisfactory dichotomies that try to capture the fundamental differences between an offensive realist state and a defensive realist state. All of them try to capture the fundamental differences in terms of the dif-ferent goals that the two realisms prescribe for states at the very abstract level: status quo versus change, relative power versus security, and absolute versus relative security, and so on. Yet, the two realisms have roughly the same assumption or conceptualization about states’ immediate and ulti-mate goals at the very abstract level: all states want security and power, and the more the better.

The fundamental difference between offensive realism and defensive realism lies elsewhere. It lies mostly in their different preference over strat-egies even when they are moving toward the same set of goals, such as power, security, and prestige.53

The Fundamental Fault Line: Preferences over Strategies

A common perception is that the differences between the two realisms lie in their different takes on the incentives that the international system provides for expansion (Taliaferro 2000–2001, 128, 134). In light of the more structure or security-driven offensive realism developed by Labs (1997) and Mearsheimer (2001), in which states’ motivation to expand is explicitly attributed to the pressure generated by the desire to survive under anarchy, it is more appropriate to argue that the two realisms differ on the different pressures and incentives that the international system provides for expansion.

Specifically, defensive realism believes that anarchy provides strong incentives for mostly defensive (or moderate and restrained) behaviors and weak pressure for offensive behaviors. Defensive approaches are the best means toward security under most circumstances (Snyder 1991, 11–12; see also Lobell 2002–2003, 169). In contrast, offensive realism believes that, in addition to defensive behavior,54 anarchy also provides strong pressure and incentives for offensive behavior. Indeed, offensive realism believes that offensive approaches often are the best means toward security: “The best defense is a good offense (Mearsheimer 2001, 36).” In other words, defensive realism sees little necessity and utility but quite a bit of danger in offensive behavior, whereas offensive realism sees much necessity and utility despite some danger in offensive behavior.

Offensive realism’s general preference for offensive behavior logically follows from its stance that maximizing relative power is the only sure


means of self-help. As Mearsheimer puts it, “The claim that states max-imize relative power is tantamount to arguing that states are disposed to think offensively toward other states (Mearsheimer 2001, 34).” While offensive realism does recognize the necessity of defensive measures against aggressors (e.g., balancing) just as defensive realism does, offensive realism places greater emphasis on offensive behaviors, even when a state faces no imminent threat. “States are . . . both offensively oriented and defensively ori-ented. They think about conquests themselves, and they balance against aggressors” (Mearsheimer 1994–1995, 12; emphasis added).55

Offensive realism’s general preference for offensive behavior follows from the fundamentally offensive implications that offensive realism draws from its recognition of the prevalence of balancing behavior and the security dilemma. Despite (implicitly) recognizing the prevalence of the balancing behavior (Mearsheimer 1994–1995, 12, 39, 45), Mearsheimer (2001) nonetheless recommends offensive states to act smartly to over-come others’ balancing effort.56 Likewise, even though a viable solution to the security dilemma is to alleviate it through moderation and coop-eration, Mearsheimer emphatically denies that the security dilemma can be alleviated (Mearsheimer 2001, 36). Instead, he explicitly advocates for escaping from the security dilemma, either by eliminating the other side in the relationship or by becoming a regional hegemon if the first option is not feasible.

Defensive realism’s general preference for defensive behavior follows from its denying that maximizing relative power is the only viable means of self-help. Contra offensive realism, defensive realism draws fundamentally defensive implications from its recognition of the prevalence of balancing behavior and the security dilemma. For defensive realism, the prevalence of balancing behavior means that a state that relentlessly pursues relative power will be perceived by other states as a threat and end up in being bal-anced (Walt 1987; see also Lobell 2002–2003, 169). Likewise, although a potential escape route from the security dilemma is to eliminate the other side in the relationship, defensive realists such as Jervis and Glaser emphatically deny that the security dilemma can be escaped and instead emphasize that the only viable solution is to alleviate the security dilemma through moderation and cooperation. For defensive realism, the secu-rity dilemma means that offensive behaviors will often be self-defeating and moderate behavior usually pays dividends (Jervis 1978, 1999; Glaser 1994/95). Thus, contra offensive realism, defensive realism does not rec-ommend offensive behavior, unless under extreme circumstances.

Overall, although the two realisms prescribe somewhat similar poli-cies when facing a clear and present threat, they recommend dramatically different policies when not facing an imminent threat. Offensive realism


recommends offensive behaviors even when not facing an imminent threat, whereas defensive realism recommends generally defensive behaviors even when facing an imminent threat, unless under extreme circumstances (see chapter 4 this volume).

A More Rigorous Dichotomy: Offensive Realist State versus Defensive Realist State

To avoid further confusion derived from the imprecise usage of vocabular-ies based on false fault lines and invalid dichotomies, I urge that the loose and inoperable dichotomies of revisionist state versus status quo state, (relative) power seeker/maximizer versus security seeker/maximizer, and absolute security-seeker versus relative security-seeker be eliminated from the vocabulary of academic debate. Instead, I will use the more rigorously defined dichotomy of offensive realist state versus defensive realist state.57

An offensive realist state is a state guided by offensive realism,58 thus heeding offensive realism’s implications for state behavior. Such a state consistently thinks of adopting and often does adopt offensive strategies against other states in order to further its goals. A defensive realist state is a state guided by defensive realism, thus heeding defensive realism’s implica-tions for state behavior. Such a state generally does not think of adopting and generally does not adopt offensive strategies against other states in order to further its goals, unless under extreme circumstances.59 Alternatively, an offensive realist state is a state that seeks security by intentionally harming others, whereas a defensive realist state is a state that does not seek security by intentionally harming others, unless under extreme circumstances.

This suggested dichotomy is explicitly based on and intended to capture the fundamental difference in preferences over strategies, or the different operational codes that the two realisms prescribe for states. Moreover, this dichotomy is explicitly centered on the overall intentions of a state, which in turn is indicated by a state’s behavior. As such, adopting this dichotomy puts the problem of intention squarely at the heart of the offensive realism and defensive realism divergence. As it becomes clear in the chapters that follow, making the problem of intention the central problématique in the offensive realism-defensive realism divergence will make the logic of both realisms more consistent, coherent, and compelling.

Furthermore, this dichotomy of offensive realist state versus defensive realist state promises to not only capture all the possible differences but to exclude those false fault lines between an offensive realist state and a defen-sive realist state at the same time (see chapter 4 for details).

For instance, this dichotomy implicitly excludes seeking material wealth (or national power) through economic growth as a potential demarcation


line between an offensive realist state and a defensive realist state. Although searching for material wealth is one of realism’s signatures, it cannot dif-ferentiate an offensive realist state from a defensive realist state. After all, economic growth is inherently about improving its citizens’ welfare, and no state can be blamed for trying to improve its citizens’ welfare through economic growth. At the same time, however, the dichotomy does allow economic sabotage to be a potential indicator of a state’s aggressive inten-tions. When a state sabotages another state’s economic growth (whether due to fear or envy), it is intentionally harming the other state. Such a state is more likely to be an offensive realist state.

Similarly, the dichotomy implicitly acknowledges that it will be more difficult to differentiate between an offensive realist state and a defensive realist state when both are facing another offensive realist state. At the same time, however, the dichotomy still allows an offensive realist state and a defensive realist state to be differentiated even when both are facing another offensive realist state. An offensive realist state will likely launch a preventive or preemptive strike against another offensive realist state, and a defensive realist state will be very reluctant to do so.

Finally, the new dichotomy also accommodates the possibility that a state may desire many goals (e.g., a world empire or regional hegemony) but nonetheless refrain from resorting to offensive actions to achieve its goals. Such a state is no different from a standard defensive realist state. Without resorting to offensive actions, its desire for its goals will remain just a fantasy and does no harm to others. In contrast, a state may simply desire peace (after all, even Hitler proclaimed that he desired peace), but it resorts to offensive actions to achieve its goals. Such a state is no different from a standard offensive realist state. By resorting to offensive actions to achieve its goal, however, it will inevitably and intentionally threaten other states.

Thus, the suggested dichotomy of an offensive realist state versus a defensive realist state will not only be more rigorous but also more opera-ble than existing dichotomies. For public consumption, the dichotomy of a malign state versus a benign state is appropriate.

Chapter Two

The Security Dilemma: A Conceptual Analysis

The security dilemma is one of the most important theoretical ideas in international relations (Wheeler and Booth 1992, 29).1 Not surprisingly, in the years since Herbert Butterfield (1950), John Herz (1950; 1966), and Robert Jervis (1976; 1978) first developed the concept, this concept has been extended and applied to “address many of the most important questions of international relations theory and security policy” (Glaser 1997, 172).

More importantly, the security dilemma is central to the logic of defensive realism (Glaser 1994–1995, 54; 2003, 406; Kydd 1997b, 116; Schweller 1996, 116).2 The security dilemma is arguably the theoretical linchpin of defensive realism because for defensive realists it is the secu-rity dilemma that makes possible genuine cooperation between states—beyond a fleeting alliance in the face of a common foe (Tang 2008a).3 But defensive realists are hardly the only scholars to utilize the concept. For offensive realists, the security dilemma makes war inevitable and rational (Copeland 2000, Chapter 2; Mearsheimer 2001, 35–36). Neoliberal scholars argue that one of the functions of international institutions is to alleviate the security dilemma (Axelrod and Keohane 1985; Weinberger 2003). Liberals claim that democratic institutions facilitate peace pre-cisely because they too alleviate the security dilemma (e.g., Dolye 1983; Lake 1992; Oneal and Russett 1999; Cederman 2001). And constructiv-ists have asserted that alleviating the security dilemma among states is one of the channels through which reshaping identity can remake anarchy (Wendt 1992; 1995).

Understood correctly, the security dilemma theory (model) and the broader spiral model constitute a powerful theory of war and peace via interaction. They capture some general dynamics leading to the outbreak of war and the maintenance of peace (i.e., by reversing or alleviating the security dilemma). The concept’s influence thus extends well beyond theory. The security dilemma (and the spiral model) have been deployed


to understand and have provided important insights into the origin of WWII (Jervis 1976, Chapter 3; Snyder 1985; Copeland 2000), the origin of the cold war (Copeland 2000; Jervis 2001; Kydd 2005, Chapters 3 and 4), the end of the cold war (Collins 1998; Kydd 2005, Chapters 7 and 8), and the outbreak of ethnic conflicts in former republics of the Soviet Union, the former Yugoslavia, and Africa (e.g., Posen 1993).4 More importantly, the security dilemma theory and the spiral model have been deployed for prescribing policies for some of the most pressing challenges in international politics, including managing arms race (Glaser 2004), designing a lasting peace to ethnic conflicts (Kaufmann 1996), and avoid-ing a possible conflict between a rising China and the United States as the reigning hegemon (Christensen 1999; 2002), to name just a few.

Despite its centrality, many areas of disagreement or confusion exist among IR theorists about the security dilemma.5 Because of its centrality in defensive realism, a coherent and systematic understanding of the security dilemma is essential for understanding defensive realism. Getting the security dilemma right also has resonance far beyond the intra- realism debate. In this chapter, I critically revisit the concept of the security dilemma and some of its most significant extensions, and advance a coher-ent and systematic restatement of the concept.6 In doing so, I lay part of the groundwork for a more rigorous and systematic statement of defensive realism.

Section 1 critically examines and builds on the writings of the three original proponents of the concept—Herbert Butterfield, John Herz, and Robert Jervis—to advance a more rigorous definition that provides a complete causal link from anarchy to the security dilemma. Section 2 derives key implications of the rigorously redefined concept, thus preparing the ground for clarifying prevalent misunderstandings. Section 3 reexam-ines several prominent extensions and expansions of the concept, showing that most of them have been unnecessary, misguided, or wrong.7 Section 4 proposes two major remedies for eliminating the areas of confusion within the existing literature and suggests several directions for future research. The chapter concludes by reemphasizing some general implications of security dilemma theory for managing international politics.

Toward a More Rigorous Definition of the Security Dilemma

Confusion regarding the security dilemma exists principally because many scholars, including the three original proponents of the concept, have defined the concept in loose ways. By critically examining and then building on the original expositions of the concept, this section provides a more rigorous definition of the concept.

the security dilemma / 35

The Security Dilemma according to Butterfield, Herz, and Jervis

According to Butterfield, the security dilemma can drive states to war even though they may not want to harm each other: “The greatest war in history can be produced without the intervention of any great criminals who might be out to do deliberate harm in the world. It could be produced between two powers, both of which were desperately anxious to avoid a conflict of any sort” (Butterfield 1951, 19–20).8 Butterfield then traced the immediate cause of the security dilemma to the “Hobbesian fear,” a fear he attributed to the “universal sin of humanity.”9 Butterfield further noted the critical role of uncertainty over others’ intentions in driving the security dilemma and the role of psychological factors (i.e., moral indigna-tion and self-righteousness) in exacerbating the security dilemma. Finally, he asserted that the security dilemma is the fundamental cause behind all human conflicts (20–22).

Butterfield emphasized six aspects of the security dilemma: (1) the ultimate source of the security dilemma is fear, which is derived from the “universal sin of humanity”; (2) the security dilemma requires uncer-tainty over others’ intentions; (3) the security dilemma is unintentional in origin—there is no intention of causing deliberate harm; (4) the security dilemma produces tragic results; (5) psychological factors can exacerbate the security dilemma; (6) the security dilemma is a critical driving force behind all human conflicts.

Note, however, that Butterfield’s attributing the ultimate source of the security dilemma to the “universal sin of humanity” is logically incompatible with his other theses that the security dilemma is unintentional in origin and that conflicts driven by the security dilemma are tragic. If we do harm against each other due to the “universal sin of humanity,” then we are programmed to harm (presumably, biologically?)—there is no exception to this rule.10 If this is the case, then there will be no real uncertainty about each other’s intentions: all of us must recognize that each of us is evil. Moreover, the unintentionality within states’ (harmful) behavior will be superficial: states are harming each other unintentionally only because each of them is (biologically) programmed to be malignant. This apparent contradiction within Butterfield’s definition is only resolved when Herz and Jervis attribute the ultimate source of the security dilemma to the anarchical structure of international politics.

John Herz, the man who originally coined the term “security dilemma,” elaborated as follows: “Groups and individuals who live alongside each other without being organized into a higher unity . . . must be . . . concerned about their security from being attacked, subjected, dominated, or annihilated by other groups and individuals. Striving to attain security


from such attacks, they are driven to acquire more and more power in order to escape the effects of the power of others. This, in turn, renders the others more insecure and compels them to prepare for the worst. Because no state can ever feel entirely secure in such a world of competing units, power competition ensures, and the vicious circle of security and power accumulation is on” (Herz 1951, 157).

And Herz went on: “Whether man is ‘by nature’ peaceful and cooperative, or aggressive and domineering, is not the question. . . . It is his uncertainty and anxiety as to his neighbors’ intentions that places man in this basic [security] dilemma, and makes the ‘homo homimi lupus’ a primary fact of the social life of man. Basically it is the mere instinct of self-preservation which, in the vicious circle [of the security dilemma], leads to competition for ever more power” (Herz 1951, 3–4). “It is one of the tragic implications of the security dilemma that mutual fear of what initially may never have existed may subsequently bring about exactly that which is feared most” (Herz 1961, 241). Although Herz (1951, 12) initially also believed the security dilemma to be all pervasive, he later explicitly rebutted Butterfield’s assertion that the security dilemma is the cause behind all human conflicts by noting that there was no security dilemma between Hitler’s Germany and other states (Herz 1961, 234n5).

Herz spelled out six aspects of the security dilemma: (1) the ultimate source of the security dilemma is anarchy—the lack of “a higher unity”; (2) an immediate cause of the security dilemma is states’ uncertainty and fears about each other’s intentions to do harm under anarchy; (3) states’ means of self-help—trying to escape from the security dilemma by accu-mulating more and more power—generates a cycle of power competition; (4) states’ attempt to escape from the security dilemma by accumulating more and more power may not increase their security at all, becoming self-defeating and even tragic; (5) the security dilemma can cause war, but is not the cause of all wars; (6) the dynamic of the security dilemma is a self-reinforcing “vicious cycle.”11

Herz’s original exposition on the security dilemma was not explicit on whether the security dilemma must be unintentional in origin. Nonetheless, Herz’s exposition implies that he must also believe the security dilemma is unintentional in origin. Herz asserted that the security dilemma arises from states’ accumulation of more and more power for their own security due to fear and uncertainty about other states’ intentions under anarchy. Moreover, Herz implied that the presence of evil intentions nullifies the security dilemma by explicitly asserting that there was no security dilemma between Nazi Germany and other states. His emphasis on the paradoxical and self-defeating (thus tragic) nature of seeking security under anarchy also suggests that the security dilemma was unintentional in origin for him.12


Robert Jervis has done the most to bring the security dilemma into mainstream (structural) international relations theories, but his writings arguably contain definitions no less casual than those provided by Butterfield and Herz. Indeed, Jervis’s writings do not contain a systematic definition of the concept; rather, they treat the concept briefly and inconsistently.

In various places, Jervis defined the security dilemma as “these unintended and undesired consequences of actions meant to be defensive” (Jervis 1976, 66); “many of the means by which a state tries to increase its security decrease the security of others” (Jervis 1978, 169); “one state’s gain in security often inadvertently threatens others” (Jervis 1978, 170); and “the ways in which the attempt by one state to increase its security have the effect (often unintended and unforeseen) of decreasing the security of others” (Jervis 1999, 49; see also Jervis 1982, 358; 1988, 317; 2001, 36).

In other places, Jervis emphasized other aspects of the security dilemma. “The heart of the security dilemma argument is that an increase in one state’s security can make others less secure not because of misperceptions or imaged hostility, but because of the anarchic context of international relations” (Jervis 1976, 76; see also Jervis 2001, 36); “Even if they can be certain that the current intentions of other states are benign, they can neither neglect the possibility that the others will become aggressive in the future nor credibly guarantee that they themselves will remain peaceful” (Jervis 2001, 36; see also Jervis 1976, 62);13 and “When determining . . . a true security dilemma, we have to consider whether those actions were purely defensive” (Jervis 2001, 57).14 Finally, like Herz, Jervis also emphat-ically denies that the security dilemma is at the heart of all conflicts (Jervis 1976, 75, 80; 1999, 49).

Jervis’s various definitions and elaborations stressed seven aspects of the security dilemma: (1) the security dilemma is structural in origin; (2) states’ uncertainty and fears about each other’s present and future intentions is crucial for forming and maintaining the security dilemma; (3) it is caused by defensive actions, thus unintentional;15 (4) it tends to produce unintended and self-defeating results—that is, decrease in one’s own security; (5) it tends to produce unintended and tragic results—that is, war; (6) the security dilemma can cause war, but is not the cause of all wars; (7) The dynamic of the security dilemma is self-reinforcing and resembles a spiral. Finally, in addition to emphasizing these seven aspects of the security dilemma, Jervis also made another fundamental contri-bution. The severity of the security dilemma can be regulated by both material/physical factors (e.g., technology, geography) and psychological/perceptual factors (e.g., misperceptions).16

To summarize, although there have been significant overlapping areas among the definitions and elaborations by the three original inventors of


the security dilemma (see table 2.1), none provided us with a rigorous and coherent definition of the security dilemma in one place. The lack of such a definition has unfortunate consequences. In particular, it fosters twisting, stretching, and misuse of the concept, resulting in many areas of confusion and contradiction.

Toward a More Rigorous Definition

To obtain a more rigorous and complete definition of the security dilemma, I draw from and build on the definitions and elaborations given by all three inventors of the concept. I perform four tasks: (1) retaining common elements that were explicitly or implicitly spelled out by all three; (2) adding elements that were only recognized by only one or two of them but have now been recognized as an integral part of the overall under-standing (e.g., Jervis’s discussion of the regulators of the security dilemma); (3) eliminating inconsistencies; and (4) adding precision.

Table 2.1 The security dilemma according to Butterfield, Herz, and Jervis

Aspects of the security dilemma

Butterfield(1951, 1960)

Herz(1950, 1951, 1966)

Jervis(1976, 1978, 1982, 1999, 2001)

The ultimate source is anarchy

No Yes Yes

Uncertainty over others’ intentions

Yes Yes Yes

Fear about each other Yes Yes YesLack of malign intentions

Yes, but inconsistent

Yes Yes, but inconsistent

Power competition Not explicit Yes YesSpiral-like situation Not explicit Yes YesUnintended tragic results

Yes Yes Yes

Unintended (and partially) self-defeating results

Not explicit Yes Yes

Regulators of the security dilemma

Only psychological factors were emphasized

Not emphasized

Both material and psychological factors were emphasized

Universal or conditional

Universal Conditional Conditional

An important cause of war?

Yes Yes Not explicit, although close to a Yes

The cause of all wars? Yes No No


I now define the security dilemma as follows: Under a condition of anarchy, two states are defensive realist states. The two states, however, cannot be sure of each other’s present or future intentions (because they live under anarchy). As a result, each tends to fear that the other may be or may become a predator.17 Because both believe that power is a means toward security, both seek to accumulate more and more power. Because even purely defensive capability will inevitably contain some offensive capability,18 many of the measures adopted by one side for its own security can often threaten, or be perceived as threatening, the security of the other side, even if both sides merely want to defend their security. Consequently, the other side is likely to take countermeasures against those defensive measures. The interaction of these measures and countermeasures tends to reinforce their fears and uncertainties about each other’s intentions, leading to a vicious cycle in which each accumulates more power without necessarily making itself more secure, through a self-reinforcing or posi-tive feedback mechanism. This vicious cycle can also lead to unnecessary conflicts—threats of war or war. The severity of the security dilemma can be regulated by both material factors and psychological factors. Because this definition is derived from the original definitions and elaborations of the security dilemma by Butterfield, Herz, and Jervis, I shall call this definition as “the BHJ formulation” hereafter.19

Defined as such, the security dilemma has at least eight major aspects.

1. The ultimate source of the security dilemma is the anarchic nature of international politics.20

2. Under anarchy, states cannot be certain about each other’s present and future intentions. As a result, states tend to fear each other (or the possibility that the other side may be a predator).21

3. The security dilemma is unintentional in origin: a genuine security dilemma can exist only between two defensive realist states (i.e., the two states merely want security without intending to threaten the other).

4. Because of the uncertainty about each other’s intentions and fear, states resort to the accumulation of power or capabilities as a means of defense, and these capabilities inevitably contain some offensive capabilities.

5. The dynamics of the security dilemma is self-reinforcing and often leads to (unintended and bad) spiral-like situations, such as the worsening of relationships and arms race.

6. The dynamics of the security dilemma tends to make some measures for increasing security—for example, accumulating unnecessary


offensive capabilities—self-defeating, creating more power but less security.22

7. The vicious cycle derived from the security dilemma can lead to tragic results, such as unnecessary or avoidable wars.

8. The severity of the security dilemma can be regulated by both material factors and psychological factors.

Among the eight aspects, three aspects are essential: anarchy (which leads to uncertainty, fear, and the need for self-help for survival or security); lack of malign intentions on both sides; and some accumulation of power (including offensive capability).23 Other aspects are either conse-quences or regulators of the security dilemma, and they are neither suffi-cient nor necessary for the rise and continuation of the security dilemma. Put differently, the three essential aspects are what make a situation a gen-uine security dilemma, and the other aspects cannot make a situation a genuine security dilemma, however powerfully they may operate, if the three essential aspects do not operate.

With the more rigorous definition, it becomes evident that the com-plete causal link from anarchy to the security dilemma and then to war is rather lengthy and by no means straightforward, and it can be captured as follows: Anarchy generates uncertainty; uncertainty leads to fear; fear then leads to power competition; power competition activates a (dormant) security dilemma; and the activated security dilemma leads to war through a spiral (figure 2.1).24

Immediate Implications of the New Definition

With the security dilemma rigorously defined (as above), six critical points must be stressed at the outset.

First, the upper part of figure 2.1 speaks about the fundamental and proximate causes of the security dilemma, whereas the lower part of this figure speaks about the potential outcomes that can be produced by the secu-rity dilemma, and these two parts must be understood separately. Although the security dilemma may produce certain types of outcomes, not all those types of outcomes are produced by the security dilemma. Thus, while the security dilemma can produce unintended and self-defeating results, not all such results are produced by the security dilemma.25 Likewise, while the security dilemma can produce (bad) spiral-like situations (e.g., a dete-rioration in relations), not all spirals are caused by the security dilemma.26 Finally, and most critically, while the security dilemma can potentially lead to war, not all wars are caused by the security dilemma (Jervis 1976, 75; Jervis 1999, 49; Kydd 1997a, 373; Kydd 1997b; Schweller 1996).


Second, we should clearly differentiate the causal factors that give birth to the security dilemma (anarchy, uncertainty and fear, and some accu-mulation of power) from the material and psychological regulators of the security dilemma. Although regulators regulate the severity of the security dilemma, they can neither give birth to nor are they necessary for maintain-ing the security dilemma (Kydd 1997a; see also Jervis 1976, 62–67; Glaser 1992, 507–508). Thus, just because some regulators (e.g., misperceptions) are exacerbating a (spiral-like) situation, it does not mean that the situation is a security dilemma. By the same token, the absence of certain regulators does not mean that the situation is not a security dilemma.

Third, we should clearly differentiate the causal factors that give birth to the security dilemma from the potential intermediary outcomes that

Anarchy

Need for self-help towardsurvival or security

through

Uncertainty and fear

Self-help throughpower competition

The Security Dilemma

Actions & reactions Material andpsychological regulators

A spiral

Unintended consequences:War or threat of war (as a

tragedy)

Unintended consequences:Partially self-defeating results(more power but less security)

The

cau

sal l

ink

to th

e se

curit

y di

lem

ma

Pot

entia

l con

sequ

ence

sof

the

secu

rity

dile

mm

a

Figure 2.1 The causal link from anarchy to the security dilemma and war.


are induced by the security dilemma but at the same time can come back to reinforce the security dilemma through a feedback mechanism. For instance, although an arms race can reinforce a security dilemma, the arms race is a possible outcome, not a source, of the security dilemma. Likewise, while some state behaviors can generate uncertainty and fear in other states, which can then exacerbate the security dilemma, this uncer-tainty and fear should not be confused with the original uncertainty and fear dictated by anarchy. Feedbacks are not causes.

Fourth, and most critically, three essential aspects (anarchy, lack of malign intentions, and some accumulation of power) are absolutely necessary for a genuine security dilemma to exist. Neglecting any one of these essential aspects will inevitably lead to error. One cannot identify a particular situation as a security dilemma just because that situation possesses several unessential aspects of the security dilemma: A situation is a security dilemma only if it has all three essential aspects of the security dilemma. If one is allowed to label a situation that has some of the unes-sential aspects yet lacks the essential aspects of a security dilemma as a security dilemma, then the security dilemma is “in danger of becoming a meaningless and ambiguous term associated with any deterioration in relations” (Collins 2000, 24).

A fifth point follows immediately from the fourth point. Because anarchy and some accumulation of power are usually, if not always, present, this leaves lack of malign intentions as the most critical ingredient for identifying whether a situation is a genuine security dilemma.27 When one or two sides in a situation is malign (i.e., intentionally threatening), a situation is not a genuine security dilemma even if it has all other aspects of a security dilemma such as spiral and unintended consequences (Collins 2000, 22–25, 175–176; Jervis 2001, 57).

Finally, exactly because the security dilemma contains eight aspects, the security dilemma as a concept and a theory is both extremely inclusive and restrictive. The security dilemma can accommodate many situations and factors (e.g., geography, hatred, nationalism), but certain situations and factors simply do not belong to the secu-rity dilemma either as a concept or a theory of state interaction (e.g., malign intentions).

As will become clear below, the fundamental reason there have been so many misguided expansions and extensions of the security dilemma is that their proponents neglected some of the principal implications outlined above.

Before I examine those misguided extensions and expansions in detail, however, let me first elaborate on several critical aspects of the security


dilemma that were either neglected or were only implicit in the BHJ formulation of the concept.

Further Elaborations

Conflict of Interest and the Security Dilemma

With the security dilemma defined rigorously, we can now address the relationship between the security dilemma and conflict of interest—a relationship neglected by the three originators of the security dilemma.28 This significant omission provides the ground for the serious charge that defensive realism depends on misunderstanding and misperception to drive conflicts among states (Schweller 1996, 118–119).29 This argument is potentially damaging to defensive realism, which, as a strain of real-ism, must assume that international politics is fundamentally conflictual: real conflict of interest among states is the norm, not the exception. As a linchpin concept of defensive realism, the theory of security dilemma must therefore also start with the assumption that there are often genuine conflicts of interest among states.

Conflict of interest is not actual (violent) conflict. Conflict of interest merely means there is some divergence between two states’ interests—that is, they want different things, or that they cannot have the same thing at the same time.

What differentiates defensive realism from offensive realism is that defensive realism recognizes that (conf lict of ) interest has both a sub-jective side and an objective side (Wolfers 1952), and that conf lict of interest can be both reconcilable and irreconcilable.30 And these two dimensions make the possible combinations of conf lict of interest quite complex (see figure 2.2 for details). As a result, the relationship between conf lict of interest and the security dilemma is also quite complex.

Let’s first deal with the situation in which there is no objective (i.e., genuine) conflict of interest but there is subjective conflict of interest between two states.31 Under this situation, there are two possible scenarios (figure 2.2a).

The first scenario is that although the two states mistakenly believe there is an objective conflict of interest between them, they also correctly believe that any potential conflict of interest is genuinely reconcilable. Under this scenario, the security dilemma may operate, but it will be an extremely benign kind. Actual conflict under this scenario is extremely unlikely, partly because the two sides perceive the situation at least par-tially correctly (i.e., misperception here is not severe).


Figure 2.2 Conflict of interest and the security dilemma.1 (a) When conflict of interest is objectively illusory but subjectively genuine. (b) When conflict of interest is objectively genuine.Note : 1 Because this figure is also essential for understanding the relationship between incompatibility and the security dilemma, I also label each situation with compatibility or incompatibility. Boulding (1959, 130–131) seems to have focused exclusively on the subjective side of this table. Jervis recognized that (security) interest has both an objective side and a subjective side, but did not elaborate on the implication of this duality for the applicability of the security dilemma (Jervis 1976, 76n36, 76n37).

Objectively reconcilable,Subjectively reconcilable.

A situation of genuinecompatibility.

Actual conflict is highlyunlikely.

The security dilemma applies,but may well remain dormant.

Objectively reconcilable,Subjectively irreconcilable.

A situation of illusoryincompatibility.

Actual conflict is not inevitable, but not unlikely.

The security dilemma may or may not apply.

Objectively irreconcilable,Subjectively reconcilable.

A situation of illusory compatibility.

Actual conflict is almost inevitable.

The security dilemma usuallydoes not apply.

Objectively irreconcilable,Subjectively irreconcilable.

A situation of genuine incompatibility.

Actual conflict is almost inevitable.

The security dilemma generallydoes not apply.

Subjective

Obj

ectiv

e

No objective conflict of interest,Subjectively reconcilable.

A situation of genuine compatibility.

Actual conflict is highlyunlikely.

The security dilemma may apply.

No objective conflict of interest,Subjectively irreconcilable.

A situation of illusory incompatibility.

Actual conflict is likely but highlyavoidable.

The security dilemma may ormay not apply.

Subjective(a)

(b)

The second scenario is that although no objective conflict of interest exists between two states, one or both states believe that the conflict of interest between them is genuinely irreconcilable. Under this scenario, the security dilemma may or may not apply, and it critically depends on whether malign intentions are present.

If the subjective irreconcilability of interest between the two states is due to one or both sides intentionally defining their interest in an


expansive way,32 then one or both sides are already malignant and the secu-rity dilemma does not apply. If this is the case, actual conflict is highly likely. If, however, the subjective irreconcilability of interest is due to one or both sides’ mistaken (thus unintentional) belief that their interest is incompatible, then the security dilemma will apply. Under this scenario, actual conflict is likely but most tragic because it is highly avoidable, and misperception will have to play a prominent role if actual conflict is to result. (Apparently, misperception is already severe here.)33

When there is objective conflict of interest between two states, the situation gets even more complex. Here, we have four possible scenarios (figure 2.2b).

The first scenario (the upper left quadrant of figure 2.2b) is one in which the conflict of interest is both objectively and subjectively recon-cilable. The security dilemma operates under this scenario but may well remain dormant. Both sides correctly recognize that the conflict of interest between them is reconcilable and neither harbors malign intentions. As a result, the power competition necessary for activating the dormant security dilemma does not have to take place and the security dilemma can remain dormant. Under this scenario, actual conflict is highly unlikely, most crit-ically because both sides recognize the situation correctly. Nonetheless, actual conflict is still possible because the security dilemma can be easily activated by one or both sides’ missteps and the security dilemma can then propel the two sides into actual conflict.34

The second scenario (the upper right quadrant) is one in which the conflict of interest is objectively reconcilable but subjectively irreconcilable. Again, the security dilemma applies only in some situations. When one or both states mistakenly believe that their interests are irreconcilable—the irreconcilability or incompatibility is due to misperception, thus illusory—but neither harbor malign intentions, a genuine security dilemma operates. Here, the subjective irreconcilability may be present at the very beginning or be the product of one or both sides’ mistaken belief that is induced by a security dilemma spiral.35 In this situation, actual conflict is not unlikely but highly avoidable through compromise (i.e., cooperation), and much depends on whether the two sides can recognize that their interests are not genuinely incompatible and take steps to signal and learn each other’s (benign) intentions. Again, misperceptions must be quite severe if this situation is going to end in actual conflict.

The subjective irreconcilability, however, can also be because one or both sides intentionally define their interest in an expansive manner, thus intentionally creating the subjective irreconcilability. If this is the case, one side’s intentions may be already malignant and the security


dilemma may not apply. Actual conflict is highly likely in this kind of situation.36

Thus, under the second scenario, actual conflict is neither always inevitable nor always avoidable. Most important, actual conflict can be driven by a genuine security dilemma spiral or by one or both sides’ malign intentions.

The third scenario (the lower left quadrant) is one in which the conflict of interest is objectively irreconcilable but subjectively reconcil-able, a case of illusory compatibility. Under this scenario, misperceptions again loom large, but it goes the exact opposite way of the conventional formulation. One side is already malign, but the other side mistakenly (perhaps only temporarily) believes that their interests are reconcilable and conflict can be avoided.37 Under this scenario, despite the presence of profound misperceptions, the security dilemma usually does not apply and actual conflict is almost inevitable, because one side is likely already malignant.

The fourth scenario (the lower right quadrant) is one in which the conflict of interest is both objectively and subjectively irreconcilable. The security dilemma generally does not apply because one or both sides are likely already malignant and both sides know it. Under this scenario, actual conflict is almost inevitable (unless one side yields to the other’s demands).38

To summarize, the relationship between conflict of interest and the security dilemma is quite complex. Nonetheless, several general observations can be made.

Foremost, the security dilemma is indeed compatible with genuine conflict of interest, although the existence of genuine conflict of interest between two states is neither necessary nor sufficient for a security dilemma to exist between them. Thus, admitting that the security dilemma does not mandate defensive realism to rely solely on misperceptions to drive all actual conflicts in international politics, as Schweller charged. Indeed, defensive realism will readily admit that genuine conflicts of interest often play an important role in driving states into arms races and actual con-flicts, but then emphasize that this is so partly because a security dilemma with genuine conflicts of interest will be far more difficult to manage than a security dilemma without genuine conflicts of interest.

Second, the theory of security dilemma underscores the possibility that two states can end up in actual conflict even if there is no objective conflict of interest between them or the conflict of interest between them is only subjectively irreconcilable but objectively (i.e., genuinely) reconcilable. In other words, defensive realism recognizes the possibility that misperception can drive states into actual conflicts. In contrast, offensive realism posits


that all actual conflicts are driven by genuinely irreconcilable conflicts of interest, thus essentially denying any role for misperception in driving conflicts. For offensive realism, actual conflicts are not tragic in the strict sense.

Third, defensive realism does not believe that all actual conflicts are inevitable, because not all conflicts of interest in international politics are genuine and irreconcilable. Defensive realism thus recognizes that at least some actual conflicts are unnecessary—thus avoidable and tragic— especially when the conflict of interest is either illusory or reconcil-able despite being genuine. When the conflict of interest between two states is illusory or reconcilable despite being genuine, and neither side harbors malign intention, the two states can avoid actual conflict through “cooperation under the security dilemma.”

Finally, despite emphasizing the possibility (or danger) that the secu-rity dilemma and misperceptions can drive states into conflict, defen-sive realism does not necessarily deny the possibility that aggressors (i.e., offensive realist states) have been responsible for most actual conflicts in international politics. Defensive realism merely argues that some, but not all, actual conflicts in international conflicts were mostly driven by the security dilemma and misperceptions (Jervis 1976, 75; Jervis 1999, 49; see also chapter 3 in this volume).

The Security Dilemma Is Conditional, Not Universal

Many scholars hold that the security is an intractable feature of interna-tional politics. As Trachtenberg (2003, 191) puts it, “the logic of the security dilemma . . . applies to all states in the system.” For Taliaferro (2000–2001, 131, 136), defensive realism posits that the security dilemma is the ineluc-table consequence of anarchy, while offensive realism denies its existence. Others, including Glaser (1997) and Wheeler and Booth (1992, 30), also hold such a conviction.

With the security dilemma defined rigorously, it becomes clear that the existence of the security dilemma is conditional, not absolute. The security dilemma is not an inherent property of anarchy (Wendt 1992, 401–402, 407; Collins 2000, 174).

Because the security dilemma requires lack of malign intentions to operate, when one or two states are intentionally threatening each other, there can be no real security dilemma between them (Schweller 1996, 117; see also Snyder 2002, 156).39

Indeed, between an offensive realist state and a defensive realist state, because the offensive realist state intentionally threatens the defensive realist state, there is no real security dilemma, and only a false security


dilemma exists between them. Only the defensive realist state may (mistakenly) perceive the situation as a genuine security dilemma. In this situation, however, the defensive realist state does itself more harm than good if it imagines a (false) security dilemma. The defensive realist state’s concern for the security dilemma may provide the offensive realist state with the much coveted opportunity for advancing its offensive goals (Glaser 1997, 191).

Thus, contrary to the impression created and held by many, the logic of the security dilemma does not apply to all states (Herz 1961, 234n5; Jervis 1976, 75; Jervis 1999, 49; Snyder 1984, 462n2). The security dilemma only exists between two defensive realist states that cannot be sure of each other’s (benign) intentions. The security dilemma is conditional, not absolute. It is not an inherent property of anarchy.40

The Security Dilemma Is Structural, Not Perceptual, in Origin

The BHJ formulation of the security dilemma explicitly stipulates that the security dilemma is structural in origin. Only a competitive self-help anarchy can give birth to the security dilemma.41 Yet, because psycho-logical factors such as uncertainty and fear play prominent roles in link-ing anarchy with the security dilemma, and psychological factors such as misperceptions and ethnocentrism can, and often do, play significant roles in driving the security dilemma, many scholars tend to maintain that the ultimate source of the security dilemma is not anarchy, but rather psychological factors.

For instance, Schweller (1996, 117–119) maintains that defensive realism (which he construes as Waltzian neorealism) relies on mispercep-tions such as “misplaced fear” rather than structure to drive the secu-rity dilemma and conflict in international politics. Schweller is wrong in at least two aspects.42 First, Schweller’s argument fails to recognize that “uncertainty about the aims of others is inherent in structural anarchy” (Snyder 1984, 462n2). For structural realism, uncertainty about others’ intentions is inherent under anarchy, thus structural in origin. The ultimate source of uncertainty and fear—as proximate causes of the security dilemma that link anarchy with the security dilemma—is not, as Schweller suggests, psychology alone, rather it is both structure and psychology. If this be so, then Waltz did rely on the structure, not mis-perceptions (or psychological factors), to drive conflicts in international politics.

Second, Schweller called the fear about others “misplaced,” thus a type of misperception (Schweller 1996, 117; emphasis added). By doing so, he failed to grasp that some amount of fear under anarchy is not “misplaced,”


but mandatory. Under anarchy, there is no 911 to call, and every state has to fend for itself. Because uncertainty about others’ intentions is inherent under anarchy, a state must fear the possibility that others may hurt it right now or in the future. Most critically, Schweller mistakenly believes that the ultimate source of uncertainty and fear—as proximate causes of the security dilemma that link anarchy with the security dilemma—is psy-chology alone rather than both structure and psychology. As a result, he con-fused uncertainty and fear as proximate causes of the security dilemma with uncertainty and fear as possible causes of misperceptions or simply misperceptions.

Yet, according to structural realism and the BHJ formulation, although uncertainty and fear do play intermediate roles in linking anarchy with the security dilemma and they can also exacerbate misperceptions among states, the ultimate source of uncertainty over others’ intentions, and thus over fear, is the structure. As Jervis (1976, 76) puts it, “One state’s secu-rity can make others less secure not because of misperceptions or imaged hostility, but because of the anarchic context of international relations.”43

By making the source of fear as purely psychological, Schweller thus committed the mistake of “over-psychologizing,” which Jervis (1976, 75) warned against many years ago.

Lack of Malign Intentions: Cannot Drop or Replace It

The notion that a lack of malign intentions is essential to the existence of security dilemma is explicit in the BHJ formulation. But decoding other states’ intentions is inherently difficult under anarchy. Consequently, when determining whether a situation is a genuine security dilemma, many scholars seek to escape the intractability of gauging others’ intentions by simply dropping the lack of malign intentions from their definition of the security dilemma altogether, or employing two seemingly equally valid but more tractable concepts (illusory incompatibility and security-seeking motive). Unfortunately, these attempts to escape from the central problem of intentions are invalid and misleading. This central problem cannot be escaped from or ignored (Tang 2008a; see also Rathbun 2007).

Dropping the Lack of Malign IntentionA recent instance of dropping the lack of malign intention from the definition of the security dilemma is provided by Booth and Wheeler (2008, 4–5). While an improvement over their previous attempt (Wheeler and Booth 1992), their new definition remains seriously flawed because the lack of malign intention is still nowhere to be found. Instead, Booth and Wheeler emphasized weapons, uncertainty (associated with other states’


weapons), and fear as the three essential ingredients. Because uncertainty and fear are essential derivates of anarchy, according to the structural ori-gin of the security dilemma, and weapons are capabilities in the original BHJ formulation, Booth and Wheeler’s definition has only two essential ingredients, neither of which is lack of malign intentions.

Booth and Wheeler overlook the fact that A’s uncertainty about B’s intentions does not equal the lack of malign intentions on A’s part or B’s part. One can be malign and still uncertain about the other side’s inten-tions. Alternatively, one can be uncertain about the other side’s intention despite the other side being really malignant. In both cases, there is no security dilemma. Just because Britain and France were unsure of Hitler’s intention before 1936 does not mean that there was a security dilemma between Hitler’s Germany and Britain/France. Likewise, just because Hitler was not sure of Britain’s and France’s intention and resolve to resist his evil designs does not mean that there was a security dilemma between Hitler and Britain/France. Between Hitler and Britain/France, there was only a security threat.

Illusory IncompatibilityAfter Jervis introduced Boulding’s dichotomy of real versus illusory incompatibility (of security interest) into the literature of the security dilemma and the spiral model (Boulding 1959, 130; Jervis 1976, 76, 80; Jervis 1978, 175–176),44 the dichotomy has gained popularity. For many, the logic seems straightforward. When there is only illusory incompatibil-ity, there is a lack of malign intentions, thus a genuine security dilemma, and vice versa.

For instance, after singling out lack of malign intention as one of the three key characteristics of the security dilemma, Collins then asserts that “determining whether the incompatibility is real or illusory is critical in determining the applicability of the security dilemma.” For Collins, the reason is purely instrumental. It is simply easier to determine whether there is illusory incompatibility than to determine whether there is malign intent (Collins 2000, 18–19, 22–25, 175–176; quotation from 22).45

Replacing lack of malign intentions with illusory incompatibility is invalid for two reasons. The first reason is that ontological reason must take priority over instrumental reason when deciding how to capture a situation. Just because one concept seems to be more operable than another does not mean that the former is more suitable for labeling a particular situation than the latter. Whether a concept is a fitting label for a situation must foremost be determined by whether it can accurately capture the essence of the situation. In this case, while determining whether there is illusory incompatibility may be more operable than determining whether


there is a lack of malign intentions, illusory incompatibility is an ill-suited concept for determining the applicability of the security dilemma, because illusory incompatibility does not necessarily indicate a lack of malign intentions.

The reason is that security interest has both an objective side and a subjective side, and this fact essentially makes equating illusory incompat-ibility with lack of malign intentions inoperable (see the section on conflict of interest and the security dilemma and figure 2.2 for details).46 Briefly, the security dilemma surely does not apply in a situation of illusory incom-patibility that is caused by one or both sides intentionally defining their security interests in an expansive way even though their objective security interests are compatible. This combination of subjective incompatibility with objective compatibility is a situation of illusory incompatibility, yet there is no genuine security dilemma here because the “illusory incompati-bility” here is due to one or both sides intentionally defining their security interest in an expansive way. One or both sides may be already malig-nant.47 Hence, the security dilemma applies to only a subset of situations with illusory incompatibility because illusory incompatibility may be due to malign intentions.

The second and equally important reason illusory incompatibility cannot replace lack of malign intentions when trying to determine whether a situation is a security dilemma is that subjective “illusory incompatibility” (assuming objective compatibility) can also be the result of security dilemma dynamics.

According to the BHJ formulation, a security dilemma can potentially lead two states that are originally benign to view each other as so threaten-ing that they come to believe that their security requirements are incom-patible. At this stage, a severe illusory (subjective) incompatibility exists between the two states. Yet, this illusory incompatibility is the result of the dynamics of the security dilemma,48 rather than a reflection of the two states’ lack of malign intention at the very beginning of their interactions. As such, the security dilemma may or may not apply here, despite the pres-ence of subjective “illusory incompatibility”.

Overall, while identifying illusory incompatibility may indeed be easier than fathoming states’ intentions, identifying illusory incompatibility does not help that much if we wish to determine whether a situation is a genuine security dilemma.

Security-seeking MotiveSecurity-seeking motive is another concept that has been suggested for determining whether a situation is a security dilemma. The basic notion, introduced by Glaser (1992) and then taken up by Taliaferro (2000–2001)


and Roe (2001, 2004), is that when states’ motives are security-seeking (i.e., when states’ behaviors are driven by security interests), the security dilemma applies. In contrast, when states’ motives are non-security- seeking (i.e., when states’ behaviors are driven by non-security interests), security dilemma dos not apply, or at least its applicability weakens.

To make identifying security-seeking motive as the central task for determining whether a situation is a security dilemma is also invalid. To do so, one basically has two options, and neither option can stand close scrutiny, because both eventually violate the BHJ formulation.

The first option is to argue, as Glaser (1992, 499–500, 502) did initially, that motives and intentions are independent of each other,49 and yet the lack of malign intentions is no longer central to the security dilemma. Thus, we are now replacing lack of malign intentions with security-seeking motive as one of the essential ingredients of a genuine security dilemma. This requires us to violate the BHJ formulation at the very beginning, and Glaser did not give any good reasons for us to do so. Moreover, as Glaser himself noted, even greedy states may be insecure (i.e., they also seek secu-rity). If so, then to make identifying security-seeking motive as the central task for determining whether a situation is a real security dilemma, we must accept that the security dilemma is compatible with both non-greedy states and greedy states and that the security dilemma is universal rather than conditional.50

The second option is to admit that lack of malign intentions is still cen-tral for determining whether a situation is a genuine security dilemma, but then to argue that security-seeking motive and lack of malign intentions are essentially equivalent.51 This seems to be the position that Glaser now holds (perhaps because the first option blatantly violates the BHJ formula-tion). Thus, while Glaser (1992) emphasized that motives and intentions are independent of each other (and we have to pay attention to both), and he still enlisted both motives and intentions in 1994 (Glaser 1994, 67), he had completely phased out “intentions” by 1997 (Glaser 1997, 2004). As a result, Glaser engineered a conceptual replacement. Determining security-seeking motives now replaces determining lack of malign intentions as the central problem for identifying a security dilemma. This, however, is too incompatible with the BHJ formulation, simply because security-seeking motive does not necessarily indicate lack of malign intentions. Security-seeking motives and benign intentions are not equivalent.

To begin with, equating security-seeking motive with lack of malign intentions is implicitly or explicitly based on the dichotomy of security-seeking state versus (relative) power-maximizing state, which can suppos-edly capture the fundamental differences between a defensive realist state and an offensive realist state. Offensive realist states are power- maximizing


states (thus malign), while defensive realist states are security-seeking states (thus benign).

Yet, as I have highlighted in the preceding chapter, this dichotomy of security-seeking versus relative power-maximizing is on shaky ground. Briefly, because every state is insecure under anarchy, every state also seeks security under anarchy. Because power remains an important foun-dation of security under anarchy for all states, and because power and security interact, “there is no possibility of drawing a sharp line between the will-to-live and the will-to-power” (Niebuhr 1960 [1932], 42).52

Consequently, it is difficult to assign this dichotomy of relative power-seeker versus security-seeker to states. This difficulty essentially makes the dichotomy of security-seeking versus relative power-maximizing unfit for differentiating states with malign intentions from states with benign intentions.

For instance, one can certainly make the case that Germany in the two world wars and Japan before WWII were at least partly seeking secu-rity. Germany wanted to escape from the predicament of its geographical location (the danger of having to fight a two-front war), while Japan wanted to make itself less dependent on imported natural resources (it wanted autarky). Thus, their behaviors were at least partly driven by security- seeking motive (Snyder 1991, 306–307).53 Yet, because they sought secu-rity by intentionally threatening others, it is hardly defensible to argue that there were security dilemmas between Nazi Germany and imperial Japan on the one side and other states on the other side (Herz 1961, 234n5), and it is at least debatable that the security dilemma played a major role in driv-ing Germany to WWI (Snyder 1985; Kydd 1997b, 153). If the behaviors of these two offensive realist states were at least partly driven by security-seeking motive, then security-seeking does not necessarily indicate lack of malign intentions. If so, then, again, security-seeking motive cannot replace lack of malign intentions as one of the three essential aspects of a genuine security dilemma.

Some (defensive) realists may counter that their definition of security-seeking is different from offensive realists’ definition of security-seeking. Their definition really means seeking security with a defensive approach (i.e., not intentionally threatening others). If so, then security-seeking motive will be equivalent to lack of malign intentions.

But this is not the way those (defensive) realists apply their definition of security-seeking motive. For instance, to bolster the case that the security-dilemma had been a major cause of war in history, Taliaferro listed many preventive wars as security-driven expansions, thus implying that they were largely driven by the security dilemma. Taliaferro seems to believe that as long as a war or an act of expansion was driven by insecurity (i.e.,


security-seeking was the motive behind it), it is a war or an expansion driven by a security dilemma (Taliaferro 2000–2001, 147–149).

Yet, of all the cases identified by Taliaferro as security-driven expansions (the Peloponnesian War, the War of Spanish Succession, the Seven Years’ War, the French Revolutionary Wars, Russia’s mobilization in the July 1914 crisis, the 1962 Sino-India War, the 1965 and 1971 India-Pakistan Wars and the 1990 Kashmir Crisis, Japan’s expansion in the 1930s, and China’s intervention in the Korean War), one can reasonably argue that only China’s intervention in the Korean War was largely driven by a secu-rity dilemma (see chapter 3 for details). In all the other cases, one or both sides in the conflict were intentionally threatening (i.e., they were not defensive), and the security dilemma cannot possibly apply in those cases if the security dilemma is meant to be the security dilemma according to the BHJ formulation.

Second and equally important, as Wolfers (1952, 485–489) recognized long ago, security has an objective side and a subjective side.54 This fact further exacerbates the problem of equating security-seeking motive with lack of malign intent. If a state defines its security in a way that requires it to intentionally threaten other states, would anyone argue that this state is still security-seeking, if security-seeking denotes lack of malign intentions? Surely no one would label pre-WWII Japan, which abhorred its lack of natural resources and thus was determined to achieve autarky by invad-ing Korea, China, and eventually the whole of East and South Asia, as a security-seeking state, if the security-seeking motive is to denote lack of malign intentions.55

Hence, to make identifying the security-seeking motive the central problem for determining whether a situation is a security dilemma is also invalid. Even if we are sure that two states are motivated by (in)security—and we can be sure, because all states under anarchy seek security—we still cannot be sure that a genuine security dilemma exists between them. One still has to make sure that they are not threatening each other intentionally.

Sorting Out the Wealth of Labels and Notions

As Jervis (1988, 318) noted, when a good idea comes along, we tend to expand and apply it widely without considering its problems and limitations. The security dilemma is one such good idea. It seems to possess some characteristics that can potentially explain many important phenomena in international politics. Thus, it is not surprising that many have expanded and applied it to “address many of the most important questions of international relations theory and security policy” (Glaser


1997, 172). Unfortunately, most of these expansions have been misguided, because they violate one or several aspects of the BHJ formulation.56

Structural versus Perceptual Security Dilemma?

Searching for possible explanations of aggressive behavior and conflicts, Jack Snyder (1985) identified four possible explanations: the structural security dilemma, the perceptual security dilemma, imperialist’s (security) dilemma, and deadlock.57

Snyder’s “structural security dilemma” depicts a security dilemma in which only material modifiers are present, while his “perceptual secu-rity dilemma” depicts a security dilemma in which both material and perceptual/psychological modifiers are present. Because states in the “structural security dilemma” and the “perceptual security dilemma” are benign, both labels do deal with genuine security dilemmas that are con-sistent with the BHJ formulation.

Nonetheless, “structural security dilemma” and “perceptual security dilemma” are really about two different types of regulators—material and psychological—of the security dilemma rather than about two dis-tinctive types of security dilemma. Snyder’s terminology generates two misperceptions.

The first misperception is that there are two fundamentally different types of security dilemma: one structural in origin, the other perceptual in origin. As I show elsewhere (Tang 2010a), this misperception provided the main intellectual foundation for Stuart Kaufman’s deeply flawed applica-tion of the security dilemma to ethnic conflict (Kaufman 1996a, 1996b). In reality, the security dilemma can only be structural in origin, although it can indeed be regulated by two types of regulators: one material or struc-tural, one perceptual or psychological.

The second misperception is that without some sort of misperception the security dilemma cannot exist (Roe 2000, 377–378). Yet, as noted above, the security dilemma is structural in origin, and neither its genera-tion nor maintenance requires misperceptions.58

Imperialist’s (Security) Dilemma?

In the 1985 paper, Snyder also coined another label, “imperialist’s secu-rity dilemma,” which he defined as a situation in which “at least one of the states in the system desires to expand, even if this entails some risk to its security.” According to Snyder, “the [imperialist] security dilemmas arise from the dynamics of limited competition over non-security interest,” and it is a “by-product of the competition over non-security interests.” “In


order to achieve its expansive political, economic, or ideological goals, the aspiring imperialist develops offensive military forces for the purpose of conquest or intimidation, and when resistance is met, a test of will and capabilities ensues. An arms race then occurs as the imperialist and its opponent both try to prove that they have the capabilities to achieve their goals.” For Snyder, the “imperialist’s security dilemma” constitutes a genu-ine security dilemma because “both competitors may prefer some compro-mise to a major war” (155–156, 165–166).

Because the imperialist’s security dilemma requires one state—the imperialist or, more precisely, an offensive realist state—to intentionally threaten other states, there can be no genuine security dilemma between it and any other state, as Herz (1961, 234n5) noted long ago. The concept of an “imperialist’s security dilemma” clearly violates the BHJ formula-tion that a security dilemma can only apply when neither side harbors malign intentions. To make his “imperialist security dilemma” stick, Snyder now merely requires the imperialist to possess some “willingness to compromise.”

Deliberate Security Dilemma?

Once we recognize that Snyder’s “imperialist’s security dilemma” is not a real security dilemma but is derived from a gross misunderstanding of the security dilemma, it becomes obvious that “deliberate security dilemma,” a similar concept developed independently by Wheeler and Booth (1992, 31, 43), is equally wrong.

Wheeler and Booth coined the label “deliberate security dilemma” to explicitly differentiate it from the unintentional or “inadvertent” form of security dilemma as originally defined by Butterfield, Herz, and Jervis (Wheeler and Booth 1992, 10). Wheeler and Booth listed two possible scenarios for “deliberate security dilemmas.”59

The first scenario arises when “a militarily status quo state adopts delib-erately ‘offensive’ strategies in order to deter another, because it sees itself in an adversarial relationship with it.” This is simply confusing a potential out-come of the security dilemma with the source of the security dilemma. The BHJ formulation explicitly allows the possibility that a security dilemma can drive two status quo states (i.e., defensive realist states) to view each other as so threatening that they adopt deliberately “offensive” strategies to deter each other (and eventually to launch preemptive or preventive war against each other). Thus, the situation depicted by Wheeler and Booth’s first scenario of “deliberate security dilemma” is not a distinctive form of security dilemma, but merely a possible outcome of the security dilemma dynamics (i.e., a deep or much exacerbated security dilemma).


The second scenario for the “deliberate security dilemma” denotes a situation in which a revisionist or revolutionary state (an offensive realist state) “adopts a posture designed to lull the target state into a false sense of security.” Such a scenario is again not a security dilemma because one side is already malignant in intentions although it tries to conceal its true (malign) intentions. The situation merely represents a cognitive challenge (and a genuine strategic threat) for the target state to correctly fathom the true (and malign) intentions of the aggressor. Snyder’s “imperialist’s security dilemma” and Wheeler and Booth’s second version of the “delib-erate security dilemma” are identical. Both capture the situation between an offensive realist state and a defensive realist state, and such a situation is not a security dilemma (Collins 2000, 10–13; 2004, 32–34).

State-induced Security Dilemma?

Alan Collins correctly recognizes that both Snyder’s “imperialist security dilemma” and Wheeler and Booth’s second type of “deliberate security dilemma” are not genuine security dilemmas because one side is already malignant. Collins nonetheless tries to salvage the enterprise of “deliberate security dilemma” by juxtaposing the first version of Wheeler and Booth’s deliberate security dilemma with a hegemonic power and then relabeling it as “state-induced security dilemma.”

For Collins, a state-induced security dilemma arises “when a status quo (but aspiring hegemonic) power pursues a deliberately aggressive policy vis-à-vis its neighbors in order to intimidate them. The aim is not to over-throw the existing status quo, but rather to consolidate it by making others too frightened to challenge it. This aggressive policy is not a precursor to war; instead it is designed to provide the state with security by requiring others to feel insecure. The state is seeking a position of hegemony and the security dilemma can arise in this situation because the other states are unlikely to be able to distinguish this approach from a revisionist state that harbors malign intent” (Collins 2000, 10–13; 2004). Collins believes that a security dilemma can still operate between an aspiring hegemon that is aggressively seeking hegemony and its neighbors if the hegemon does not seek war but merely submission from its neighbors.

This label of hegemonic state-induced security dilemma is again mis-taken. Because Collins’s formulation is a hybrid of Snyder’s “imperialist security dilemma” and Wheeler and Booth’s “deliberate security dilemma,” it suffers from many of the same flaws. In addition, Collins’ formulation suffers from three related mistakes.

First, Collins adopts a too narrow definition of malign intentions and an offensive realist state (or revisionist state) by identifying only


“overthrowing the existing status quo” as being expansionist or aggres-sive. In reality, when a state seeks hegemony by intentionally making other states feel insecure, it is already an offensive realist state, regardless of its tactical means and intermediate goals.

Indeed, seeking hegemony is exactly what Mearsheimer prescribes for offensive realist states (Mearsheimer 2001, 33–35). Moreover, Snyder explicitly argues that an imperialist state will pursue both conquest and intimidation (Snyder 1985, 165), and Mearsheimer emphasizes that an offensive realist state is not a “mindless aggressor” (Mearsheimer 2001, 37), thus also implying that intimidation is a tool of an offensive realist state. Hence, seeking to establish hegemony with intimidation, although perhaps more palatable for the victim, is still a form of intentionally threatening other states. As a matter of fact, intimidation is an espe-cially “wise” strategy when the aspiring hegemon still lacks the material power to impose its will, or it is simply too costly to impose hegemony by force, because it allows the hegemon to be viewed as not “excessively” aggressive.

Second, the adjective “state-induced” adds no descriptive value. Although the ultimate origin of the security dilemma is anarchy, the exis-tence of security dilemma requires the presence of states and interactions among states. In other words, the security dilemma requires both anarchy and state-to-state interaction (e.g., power competition, action and reac-tion) to exist. Anarchy without any interacting units within it will be an empty shell.

Finally, other than the fundamental conceptual issues, Collins’s “state-induced security dilemma” is almost impossible to operationalize. When a state is pursuing an “intentionally aggressive policy,” how can it be still called a status quo state (i.e., a defensive realist state)? Isn’t it behaving exactly like an offensive realist state? Indeed, Collins (2000, 12) came very close to admitting the inoperability of his concept: “Given how aggressive the hegemon appears, they (other states) are much more likely to interpret its action as hostile.”

Always Secure State?

In his 1992 article, Charles Glaser expanded on the spiral model and the deterrence model to understand the political consequences of mili-tary strategies. Complaining that Jervis’s elaborations on the two models focused only on states’ intentions and gave insufficient attention to states’ motives for expansion, Glaser insists that we should pay equal attention to intentions and motives (1992, 499–508).60 Glaser thus introduces a


supposedly more fine-grained differentiation of states’ motives along two dimensions. First, Glaser asks whether a state is interested in non-security expansion. A state is a greedy state if the answer is yes and a not-greedy state if the answer is no. Second, Glaser asks whether a state is interested in security-driven (or fear-driven) expansion. A state is a potentially insecure state (or insecure state for short) if the answer is yes and an always-secure state if the answer is no (Glaser 1992, 501–506). By combining these two dimensions of motive (greedy versus not-greedy, and always-secure versus insecure), Glaser claims, states can then be categorized into four types, providing the spiral model and the deterrence model with a much more fine-grained explanatory power and generating more calibrated prescrip-tions for states’ military strategies.

Glaser invented the dichotomy of greedy state versus not-greedy state to avoid the pitfalls associated with the dichotomy of revisionist state ver-sus status quo state (Glaser 1992, 501n4). This dichotomy of greedy state versus not-greedy state thus essentially recaptures states’ intentions, which has been partly captured by the dichotomy of revisionist state versus status quo state. Greedy states are malignant and not-greedy states are less likely to be so (see chapter 1 in this volume). As such, this dichotomy is at least partly compatible with the BHJ formulation.

The same, however, cannot be said for Glaser’s other dichotomy, the dichotomy of always-secure state versus insecure state. Glaser (1992, 502) defines the always-secure state as a state that “recognizes that the defender is interested only in protecting the status quo and would employ its military capabilities only in response to aggression. As a result, unlike a potential insecure state, an always-secure state is not made insecure by the deployment of capabilities that threaten its ability to protect the status quo.”

Such a situation, while not unthinkable, must be extremely rare. Because intentions can change, a state cannot be absolutely sure of the defender’s intentions at a later time, and thus it cannot be “always secure” even if it can be absolutely sure of the defender’s benign intentions now (i.e., being “secure” for now). Fundamentally, because uncertainty about the aims of others is inherent in structural anarchy and there is no central authority to enforce commitments, no state can be absolutely or always secure.61 Under anarchy, all states are insecure one way or the other; only the degree of their security varies. The always-secure state has rarely, if ever, existed. As a result, the differentiation of always-secure state and potentially insecure state is not viable.

When Glaser introduces the notion of always-secure state, he implicitly assumes that there can be certainty about intentions under anarchy, thus


nullifying much of the impact of anarchy on state behavior. By eliminat-ing uncertainty, Glaser severs the crucial link between anarchy and the security dilemma.

Summary: Six Common Mistakes

From the preceding discussions, we can detect six common errors in the “expansionist” understanding of the security dilemma: (1) omitting one or more of the three essential elements of the security dilemma, especially a lack of malign intentions; (2) replacing lack of malign intentions with some seemingly equivalent but more operable concepts (i.e., illusory incompati-bility and security-seeking motive); (3) conflating regulators of the security dilemma with the essential elements of the security dilemma, which often leads to over-psychologizing the concept; (4) conflating outcomes that can be produced by the security dilemma (and the broader phenomenon called the spiral) with essential aspects of the security dilemma; (5) severing or shortening the causal link from anarchy to the security dilemma; and (6) unnecessarily expanding the security dilemma to accommodate situations that the security dilemma model has already accommodated.

Each of the expansions and extensions of the security dilemma exam-ined committed at least one or several of the six errors (see table 2.2 for details). The net result is a general tendency to identify any situation

Table 2.2 Common errors about the security dilemma

Authors

Errors

Snyder (1985) Wheeler and Booth (1992)

Collins (2000) Glaser (1992, 1997), Glaser and Kaufman

(1998)

1: Omitting + + +2: Replacing + + +3: Conflating regulators with essential elements, including over-psychologizing the concept

+ + +

4: Conflating possible outcomes with essential elements

+ + +

5: Shortening the causal link

+ +

6: Unnecessary expansion

+ + + +


that resembles a spiral or any (unintended and usually self-defeating) outcome that seems to have been produced by a spiral with a security dilemma.

Implications for Future Research

The preceding discussion points to several critical implications for under-standing international politics with the security dilemma as an analytical tool. The most obvious implication is that if scholars want to use, not abuse, the security dilemma as an analytical tool they must remain faithful to the original BHJ definition.

To begin, the BHJ formulation of the security dilemma is very accom-modating. It can cover many of the situations that many thought could only be accommodated by expanding and extending the concept.

For instance, the BHJ formulation explicitly underscores that the severity of the security dilemma can be regulated by both material and psychological factors. As a result, there is no need for Snyder’s labels of “structural” or “perceptual” security dilemma. Likewise, the BHJ formu-lation also explicitly allows the possibility that the security dilemma can drive two defensive realist states that are originally benign toward each other to view each other as so threatening that they then deliberately adopt “offensive” military postures to deter each other and thus end up exacerbating the security dilemma between them. Consequently, there is no need for the first version of Wheeler and Booth’s “deliberate security dilemma.” Finally, the BHJ formulation can easily accommodate the situa-tion between a (regional) great power and its neighbors that cannot be sure of each other’s intentions, and there is no need for Collins’s “state-induced security dilemma.”

At the same time, however, some situations that cannot be accom-modated by the BHJ formulation simply do not belong to the security dilemma and should not be squeezed into the concept. Instead, these situ-ations should be captured by more accurate and appropriate labels. Most obviously, Snyder’s “imperialist security dilemma,” Wheeler and Booth’s second version of “deliberate security dilemma,” and Collins’s “state-in-duced security dilemma,” as defined, all depict the situation between an offensive realist state (i.e., an imperialist, an aggressive hegemon, or a pred-ator) and a defensive realist state. Such a situation is not a genuine security dilemma, and it is best captured by a label of “imperialist or expansionist threat,” from the point of view of the threat’s potential victims.

Hence, the problem with the BHJ formulation has not been that it is too rigid, but rather that it is under-specified and underdeveloped. As a result, its full potential has yet to be fully realized and appreciated, and


many have then mistakenly believed that they have to expand—or more precisely, to stretch, bend, and twist—the concept. What we urgently need is not more loose and misguided expansions and extensions of the security dilemma, but a more precise and systematic understanding of the concept.

Two tasks are most urgent toward such an understanding.

Differentiating the Security Dilemma (Theory) from the Spiral (Model)

The first and perhaps the most urgent task is to clearly differentiate the security dilemma (both as a phenomenon and a theory) from the spiral (both as a phenomenon and a model/theory).62

In his classic treatise, Jervis (1976, Chapter 3) recognizes that the process through which a security dilemma drives states to actual conflict resembles a spiral. Jervis thus put his discussion on the security dilemma under the “spiral model,” without ever being explicit whether a security dilemma and a spiral are equivalent, or whether the security dilemma theory and the spiral model are the same.63 Jervis’s unfortunate omission laid the ground for much confusion. As a result, many tend to believe that the security dilemma and the spiral are the same: the security dilemma produces spi-rals and all spirals are caused by the security dilemma. Consequently, the security dilemma theory and the spiral model are equivalent.64 In Glaser’s words, the security dilemma “provides the rational foundation for the ‘spi-ral model’ ” (Glaser 1997, 171; see also Copeland 2000, 52; Glaser 1992, 499; Jervis 1999, 49; Kydd 2005, 50; Schweller 1996, 117; Taliaferro 2000, 147–150).

This notion that the security dilemma and the spiral are the same, however, is incorrect. To begin with, a spiral is generally defined as “a continuously accelerating increase or decrease,” “a process of progres-sive deterioration,” or simply “a situation that gets worse and worse.”65 Translated into strategic language, a (upward) spiral merely denotes a sit-uation in which tension between two states is continuously increasing because the process is driven by a self-reinforcing mechanism. Such a def-inition says nothing about the nature of the forces that drive the process. This effectively means that an upward spiral can be driven by a genuine security dilemma between two defensive realist states or by a genuine irreconcilable conflict of interest between one defensive realist state and one offensive realist state, or between two offensive realist states. If this is so, then while it is wrong for offensive realists to assume that all spirals (e.g., arms races and actual conflicts) are driven by irreconcilable con-flicts of interests (e.g., Gray 1992; 1993), it is equally wrong for defensive


realists to assume that all, or even many, spirals are driven by the security dilemma.

Indeed, if we demand consistency between Jervis’s discussion of the security dilemma and the spiral and his stand that the security dilemma can only be genuine when neither side harbors malign intentions, then the spiral and the security dilemma cannot be equivalent. A spiral, compatible with both a presence of malign intentions and a lack of malign intentions, can exist between any two states. In contrast, a security dilemma, com-patible only with lack of malign intentions, can only exist between two defensive realist states. In other worlds, the spiral is universal whereas the security dilemma is conditional. Thus, the security dilemma applies to only a subset of spirals.

A clear differentiation between the security dilemma (theory) and the spiral (model) brings two critical benefits.66

First, it clarifies many areas of confusion in the application of the two concepts and allows us to apply the two concepts and the two theories more precisely. For instance, it becomes evident that some of the situations discussed in Jervis’s “Cooperation under the Security Dilemma” are not security dilemmas, but spirals driven by malign intentions. The scram-ble for colonies in the eighteenth and nineteenth centuries among the European powers, despite being a spiral, was not caused by the security dilemma, contrary to what Jervis asserted (Jervis 1978, 185–186). Rather, it was caused by expansionist ambitions among the imperialist powers (e.g., Britain, Russia, Revolutionary France)—among imperialist pow-ers there is no genuine security dilemma because they are offensive realist states that intentionally threaten one another.67 Meanwhile, the situation in which “statesmen believe that their security requires the threatening or attacking of others” has two possible scenarios: a scenario in which one or two states believe that their security requires the threatening or attacking of others at the very beginning and a scenario in which one or two states come to believe that their security requires the threatening or attacking of others because of the dynamics of security dilemma. Just as Jack Snyder did later, Jervis failed to realize that only the second scenario is a genuine (vicious or deep) security dilemma, whereas the first is not (Jervis 1978, 185; Snyder 1985).68

Second, the differentiation makes it clear that many expansions of the security dilemma were attempts to accommodate spirals rather than gen-uine security dilemmas. Recognizing this allows us to reintegrate many insights that were generated from the ill-informed expansions and exten-sions of the security dilemma into the broader spiral model. For instance, building upon Jervis (1978, 182), Glaser advanced a fundamental insight that “analyses that fail to consider political consequences risk prescribing


either too much or too little military capability, and often more important, the wrong kind, which can reduce states’ security” (Glaser 1992, 499). An equally important insight is Snyder’s thesis that neither pure threats (measures advocated by the deterrence model) nor pure concessions (mea-sures advocated by the security dilemma model) could handle a situation like 1914, because it was not a genuine security dilemma but more likely an “imperialist security dilemma,” or a “deadlock,” in Snyder’s terminol-ogy (Snyder 1985, 154–155). Both insights are critical for defensive realist states to formulate their security strategies, and they can and should be fruitfully integrated into the broader spiral model.69

Third and perhaps most crucial, the differentiation makes the security dilemma model (and the broader spiral model) truly dynamic. Essentially, the differentiation makes it apparent that a security dilemma can be trans-formed into a spiral when one or both sides become malignant (e.g., one or two sides may become so frightened that they may decide that their secu-rity now requires them to pursue aggression). In other words, whenever one or both sides within a security dilemma decide that their security now requires them to pursue aggression, the security dilemma ceases to exist. The dynamics of security dilemma is not only potentially destructive (i.e., leading to war) but also self-destructive.

A rigorous differentiation of the spiral and the security dilemma and the spiral model and the security dilemma model will enable more fine-grained understandings of international politics, thus providing more fine-grained policy recommendations for managing international politics today.

A Reversible and Graduated Continuum

After differentiating the security dilemma and the spiral, the logical next step is to make it explicit that both a security dilemma and a spiral can be better understood as a reversible and graduated continuum (figure 2.3).70

Briefly, when two defensive realist states first interact, a dormant and largely benign security dilemma exists between them. This benign secu-rity dilemma, however, can be easily exacerbated into a “vicious” or “deep” security dilemma in which both sides now significantly fear each other by the uncertainty and fear generated by the states’ behaviors (partly because anarchy prevents states from completely trusting each other’s intentions under most circumstances) and by the self-reinforcing cycle of action and counteraction.71

In a deep security dilemma, one or both sides may become so fright-ened (or provoked by the other side, objectively or subjectively) that they may decide that their security now requires them to pursue aggression.


At this stage, one side’s or both side’s intentions change from benign to malign. One or both states have metamorphosed from defensive realist states into offensive realist ones. As soon as this change occurs, the secu-rity dilemma stops operating, and a spiral takes over. A security dilemma is now transformed into a spiral.72 Here, the spiral can take two possible

A dormant security dilemma

An activated security dilemma

A much exacerbated ordeep security dilemma

Material andpsychological regulators

A spiral

Imperialist/expansionist threat: asecurity problem for one side

when the other sideturns malign

Mutual threat/deadlock: securitythreat for both sides when both

sides turn malign

Some accumulation ofoffensive power

Change of Intentions:from benign to malign

The

spi

ral m

odel

The

sec

urity

dile

mm

a m

odel

War or threat of war

Gra

y zo

ne

Actions andcounteractions

Actions and reactions;material and psychological

regulators

Figure 2.3 The continuum: From a security dilemma to a spiral.


forms: “imperialist or expansionist threat” when only one side becomes malignant; and “mutual threat or deadlock” when both sides become malignant.73

A deadlock is almost impossible to unwind because it would require both sides to change their mind-set. An expansionist threat is also diffi-cult, but not impossible, to unwind. When the expansionist state decides it no longer wants to expand, the situation is then changed into a genuine but still deep security dilemma.

The utility of this graduated continuum can be illustrated by the history of the cold war (see Appendix I for a more detailed discussion). Briefly, at the beginning of the cold war (up until the Turkey crisis in 1946), despite the fact that many American decision-makers had a generally benign image of the Soviet Union under Stalin and worried about the danger of an avoidable confrontation between the two superpowers driven by a secu-rity dilemma (Kydd 2005, Chapter 4; Weinberger 2003), there was only a spiral but no real security dilemma between the two superpowers because Stalin was a genuine expansionist. After Stalin’s death in 1953, especially after Khrushchev consolidated his power and repudiated some of Stalin’s expansionist policies, the cold war was then perhaps transformed into a genuine security dilemma. Yet, because the security dilemma emerged from a spiral of “expansionist threat,” the security dilemma inherited by Khrushchev and Eisenhower (and later, Kennedy) was very “deep.” In light of such a dynamic process, the right question to ask may not be whether the cold war (or some other conflict or conflictual relations) was a secu-rity dilemma, but rather when the cold war was and was not a security dilemma.

Understanding the Regulation of the Security Dilemma

The third task is to develop a more systematic and in-depth understanding of the regulation of the security dilemma and the spiral.74

Material RegulatorsSince Jervis, only four material regulators of the security dilemma have received sustained attention in the literature.75 I believe three additional material factors should be added to the list of material regulators, and they are: asymmetric power, external actors (allies), and concentration or mixing of ethnic groups.76

Considering that distribution of power is central for realism, it is perhaps surprising that asymmetric power has never been seriously considered in conjunction with the security dilemma.77 Because asymmetric distribution of power between two states is usually the norm and not the exception, we


should treat asymmetric distribution of power as an important material regulator of the security dilemma. Understanding how asymmetric distri-bution of power regulates the security dilemma will provide us with some much-needed insight into how to manage the relationship between a great power and its smaller neighbors,78 and the relationship between a reigning hegemon and a rising power.

Overall, the security dilemma theory suggests that in a relationship of asymmetric power the weaker party will be more sensitive, and thus more likely to overreact. As a result, to alleviate or at least contain the security dilemma, the stronger power needs to be more reassuring and self- restraining. Unfortunately, more power not only tends to lessen the incentives for the stronger state to be moderate, because the cost of conflict will be relatively lower for it, but also tends to make it less willing to be constrained because it can overcome the constraint (Kydd 2005, 39–40). Meanwhile, because the weaker state is more sensitive, and thus more likely to overact, it may take some measures of self-defense that will be deemed hostile by the stronger power. Here, one measure is perhaps particularly alarming for the stronger power: the weaker state forms a tight alliance with another great power with whom the stronger power does not enjoy a friendly relationship, thus arousing the stronger power’s fear of being contained by the weaker state and the external great power.79

Moreover, asymmetric power also has great implications for managing the relationship between a reigning hegemon and a rising state. In light of the continuing debate about how to manage their uneasy relation-ship in both the United States and China, this issue is perhaps especially pressing.80

On the one hand, because hegemony exacerbates the fear of (relative) decline and magnifies the sense of threat (Jervis 2006, 13; Tang 2005, 59–61), the hegemon may be overly sensitive to any growth of the rising power and thus become extremely difficult to be reassured.81 At the same time, the hegemon may well believe that it is so self-evidently benign and fail to recognize how its overwhelming power can appear extremely intim-idating (and potentially threatening) to the rising but still much weaker state (Jervis 2006, 13–14). The situation becomes worse if the hegemon is really an offensive realist state (i.e., an unsatisfied power like an imperialist state), as many have argued the United States was under George W. Bush (e.g., Bacevich 2002; Ikenberry 2002; Jervis 2006; Prestowitz 2003).82

On the other hand, while the rising state has many reasons to reas-sure the hegemon even if the hegemon often appears high-handed, the rising power has an extremely difficult task in satisfying the rising expecta-tions of its population and reassuring the always demanding and imposing hegemon at the same time.


Our understanding of the impact of external actors—especially allies and alliance—upon the security dilemma seems to have stagnated after Snyder’s contribution (Snyder 1984). Here, the more recent application of the spiral or the security dilemma to ethnic conflicts indicates that having real allies or even perceived support from potential allies can consider-ably exacerbate the ethnic spiral (Posen 1993, 35). For instance, Germany’s support and eventual recognition of Slovenia, Croatia, and Bosnia of the former Yugoslavia was perhaps an ill-informed policy and might have been an important accelerator of the eventual bloodshed (Roe 2004). Likewise, in the case of Moldova, Russia’s support for ethnic Russians in the Dniestr region may have emboldened ethnic Russians to eventually seek secession (Kaufman 1996b, 110).

Another potentially important material regulator, the mixing of ethnic groups, was first singled out and explored in the ethnic conflict litera-ture (Posen 1993). Mixing of ethnic groups can also regulate the severity of interstate security dilemma. The presence of one country’s major eth-nic group in another country as a minority group (e.g., ethnic Chinese in Indonesia and Malaysia, or ethnic Russians in Ukraine and Kazakhstan) may pose a very sensitive challenge for the two states in managing their relationship. One can easily imagine that real or perceived ill-treatment of the minority group by its host state can provoke a public outcry in the state in which the group is the major group. Such a dynamics can easily exacer-bate the security dilemma between the two states. Worse yet, the worsen-ing relationship between the two states will then come back to exacerbate the tension between the two ethnic groups.83

Psychological RegulatorsWhile psychological factors had long been noted to regulate the security dilemma and psychological factors have received increasing attention from students of IR, the literature on psychological factors in IR has generally proceeded separately from the literature on the security dilemma after Jervis’s groundbreaking study (Jervis 1976, 1978).84 The linking of nation-alism with the security dilemma in ethnic conflicts might have been the sole exception (Kaufman 1996a, 1996b; Roe 2001, 2004).85

With a growing and more sophisticated literature on psychological fac-tors in international politics, now it may be high time to rekindle the ties between psychological factors and the security dilemma.

To begin with, it will be extremely interesting to explore whether some well-documented mentalities, such as concern for reputation or credibil-ity (McMahon 1991; Mercer 1997; Press 2005; Tang 2005) and analog-ical reasoning (Khoong 1992) can influence the severity of the security dilemma or the spiral.86


A particularly prominent psychological factor is fear. Fear has featured prominently in realism theories (Tang 2008a), and it has long been argued to be a critical link between anarchy and the security dilemma and a key accelerator of the security dilemma. Yet, surprisingly little attention has been devoted to understanding how fear really drives state behaviors and the security dilemma, other than in Jervis’s brief discussion (Jervis 1976, 372–378). Considering how central fear is to the security dilemma theory (and to realism in general), it is perhaps time for a change.

The general aversion to loss as captured by prospect theory may be another major regulator of the security dilemma.87 This general aversion to loss may apply not only in interstate situations but also in intrastate situations. For instance, an ethnic group that traditionally enjoys a privi-leged position in a society will deem any possibility of lessening its status as fundamentally threatening. As a result, such a group has great incentives to seek secession or some kind of special arrangement for itself if the central authority breaks down. This fear of losing one’s privileged position cer-tainly has the potential to exacerbate the security dilemma to such a degree that one or both sides may come to believe that their security depends on excluding the other group through ethnic cleansing or secession.88

In addition, it will be extremely interesting to explore how the psycho-logical factors listed by Wheeler and Booth (1992, 40)—which includes ethnocentrism, doctrinal realism, ideological fundamentalism, worst-case forecasting, implicit enemy imaging—can potentially aggravate the secu-rity dilemma. For example, the ideological debate between the Soviet Union and China after Khrushchev’s denouncing Stalinism in 1956 and advocating “peaceful co-existence” with the Western camp was perhaps a major cause behind the collapse of the Soviet Union-China alliance (Haas 2005).89

Regimes and the Security DilemmaFinally, a most intriguing possibility is how domestic politics, which then can be linked to regime types, influences the security dilemma.

While the growing attention paid to the role of domestic politics in shaping state behavior has generated a distinctive literature of neoclassical realism (Rose 1998; Schweller 2003), the neoclassical realism literature has yet to make any substantial connection with the security dilemma theory. Meanwhile, other than Kydd’s brief discussion (Kydd 1997b, 129–139), the literature on democratic peace has yet to make any real connection with the security dilemma, although proponents of democratic peace theory have long argued that democratic institutions prevent war among democ-racies via alleviating the security dilemma (e.g., Lake 1992; Cederman 2001), and that modeling costly signaling within a democracy regime has


yielded some interesting hypotheses that can be readily connected with the security dilemma literature (Fearon 1994; Schultz 1998, 1999). It is time for a change.90

For instance, it is believed that the secrecy of decision-making within nondemocratic states exacerbates the security dilemma. Yet, while democ-racy’s signaling of resolve (to stand firm and fight) may be more credible because its decision-making process is more transparent (due to the pres-ence of opposition parties and a free media), will the open debate about policies in a democracy also exacerbate the security dilemma between a democracy and another state because the other state is likely to pay more attention to the rhetoric that signals hard-line positions?

Concluding Remarks

Because of the centrality of the security dilemma in defensive realism, our systematic revisiting of the security dilemma will naturally have major implications for the defensive realism approach toward managing international politics.

Defensive realism’s interpretation of the logic of the security dilemma maintains that when a security dilemma is real, it is inescapable. A security dilemma can only be mitigated but not eliminated, and the major way to alleviate the security dilemma is to pursue cooperation through reassur-ance (see chapter 5 in this volume). Moreover, under many circumstances, a security dilemma can be better mitigated if both players understand they are in a security dilemma. These two points are agreed upon by all propo-nents of defensive realism (and neoliberalism) since Herz, and they have major implications for how a defensive realist state manages its security under anarchy.

The most important implication for a defensive realist state to manage its relationship with other states is that it should not automatically assign malign intentions to another state. Doing so will automatically poison their relationship. Instead, a defensive realist state should try to learn the other state’s true intentions by signaling benign intentions to it through cooperative measures, without risking its vital security interest. If the other state reciprocates, then it is more likely to be a like-minded defensive real-ist state (Kydd 2005, Chapters 2 and 3). If so, the right policy for the two states is to cooperate and behave with moderation rather than to engage in a vicious cycle of power competition. Between two like-minded defen-sive realist states, cooperation becomes a viable means of self-help for their welfare.

If the other state responds by taking advantage of the goodwill of the defensive realist state, then it is more likely to be an offensive realist state.


However, even here the right policy is not, as many tend to believe, a turn to a simplistic and hard-line policy of arms race and deterrence. Rather, insights from the spiral model point to a more nuanced policy of simulta-neously reducing the insecurity of the offensive realist state, while making it clear that aggression will not pay and that moderation is in the interests of both parties (Glaser 1992, 1994–1995; Snyder 1985). Overall, such a policy requires the defensive realist state to assure the offensive realist state by signaling benign intentions while maintaining a robust deterrence and defense posture at the same time. This requires the defensive realist state to strike a delicate balance between the two objectives, and achieving this delicate balance will not be easy, because these two goals often pull in opposite directions. Yet, this balance is a task that a defensive realist state must face if “[its] security policy is to be expedient” (Wolfers 1952, 497). And it is here that security dilemma theory (and the broader spiral model) contributes the most to managing peace. The theory not only shows the danger of relying on a simplistic hard-line approach of arms race and deterrence but also shows the ways of achieving this delicate balance.


Chapter Three

The Security Dilemma and War Revisited

After clearing the underbrush that has come to obstruct our understanding of the concept of the security dilemma itself, we can now tackle the critical question: has the security dilemma been a major cause of war in history? Or, to put it differently, can the security dilemma really explain most, or even many, interstate wars in history? If the answer to the question is yes, what kind of lessons can we draw for managing international politics today? If the answer is no, does the security dilemma still have any signifi-cant relevance for managing international politics today? Because the secu-rity dilemma is central to defensive realism, our answers to these crucial questions will have important implications for understanding defensive realism’s approach toward managing security.

Because the security dilemma is a linchpin concept of defensive realism, defensive realists since Butterfield and Herz have persistently and vigor-ously emphasized its critical role in international politics. As a result, some seem to hold the impression that the vicious spiral of power competition derived from the dynamics of the security dilemma under anarchy has been an important, if not the most critical, cause of war among states in history (e.g. Glaser 1997; Taliaferro 2000–2001, 145–151).1 Because many defensive realists believe that defensive realism needs to be able to explain war and peace in history in order to be a grand theory of interna-tional politics, this thesis that the security dilemma has been a major cause of wars in human history, although often implicit, is undoubtedly one of defensive realism’s central tenets.

This chapter argues that this thesis is invalid. The security dilemma has not been a major cause of war in history. Building upon earlier attempts to address this central question (Kydd 1997b; Schweller 1996; Snyder 1985), I advance a (hopefully) definitive explanation why the security dilemma has not been a major cause of war in history. I also underscore three major causes behind the incorrect impression that the security dilemma has been a major cause of war in history. Yet, I also emphasize that just because


the security dilemma has not been a major cause of war in history does not mean that it is not important for understanding international poli-tics today. In fact, I contend that precisely because the security dilemma has not been a major cause of war in history, it is a critical instrument for understanding international politics today.

The rest of the chapter is divided into five major sections. Section 1 reviews defensive realism’s theoretical and empirical arguments for the thesis that the security dilemma has been a major cause of war in history. Section 2 critically reviews two recent challenges to defensive realism’s thesis, underscoring why neither challenge was able to fundamentally undermine the thesis that the security dilemma has been a major cause of war in history. Section 3 and section 4 advance an empirical/normative explanation and a social evolutionary explanation, respectively, showing why the security dilemma has not been a major cause of war. Section 5 underscores three major causes behind the incorrect thesis that the secu-rity dilemma has been a major cause of war in history. A brief conclusion follows.

The Security Dilemma as a Major Cause of War: The Pros

To resolve the debate whether the security dilemma has been a major cause of war in history, let us be clear about what defensive realists have argued so far, both theoretically and empirically.

What Defensive Realists Have (Not) Argued Theoretically

Perhaps surprisingly, other than Butterfield who had argued that the security dilemma lies behind all human conflicts (Butterfield 1951, 20), none of the major proponents of the security dilemma have explicitly argued that the security dilemma has been a major cause of war. Nor have they advanced a coherent argument why the security dilemma should have been a major cause of war in history.2

To begin with, Herz merely stated that: “Ultimately, somewhere, conflicts caused by the security dilemma are bounded to emerge among political units of power” (Herz, 1950, 158). Such a statement says nothing explicit about just how many wars the dynamics of the security dilemma had actually caused in human history. One thing is for sure, Herz certainly did not believe that the security dilemma lies behind all wars, as Butterfield believed, because Herz pointedly noted that there was no security dilemma between Hitler’s Germany and other states (Herz 1966, 234n5).

Likewise, Jervis also said nothing explicit about how many wars the security dilemma had actually caused in history, and his stand on this

the security dilemma and war revisited / 75

question has been inconsistent at best. In some places, Jervis asserted that “in many cases, it is the interactive process among states that generates conflict,” and “although other motives such as greed, glory, and honor come into play, much of international politics is ultimately driven by fear” (Jervis 1999, 49; 2001, 36), and both statements seem to imply that the security dilemma must have caused a significant number of wars. In other places, however, Jervis merely stated that the security dilemma makes war likely “even in the extreme case in which all states would like to freeze the status quo,” and that “the security dilemma can not only create conflicts and tensions but also provide the dynamics triggering war” (Jervis 1978, 167; 1976, 67). Still in other places, Jervis stressed that “defensive realists believe that only in a subset (size unspecified) of situations is conflict unnecessary” and that “the security dilemma can only apply to a subset of cases” (Jervis 1999, 49), thus clearly indicating that he does not believe that all wars were caused by the security dilemma and is agnostic about how many wars the security dilemma has actually caused in history.

Similarly, Glaser (1992; 1994; 1997) too has said nothing explicit about why the security dilemma has been a major cause of war in his-tory, although he believes that the security dilemma merely requires security-seeking motive to operate, thus implying that it is omnipresent (see chapter 2 in this volume).3

Finally, although the title of Waltz’s 1988 paper was “the Origins of War in Neorealist Theory” and he did cite the security dilemma as an important link between anarchy and war, Waltz too merely implied that the security dilemma is one of the systemic level factors that permit the outbreak of war and said nothing explicit about whether the security dilemma has been a major cause of war. For Waltz, the security dilemma is useful for denying that “the innate lust for power constitutes a sufficient cause of war in the absence of any other,” and then asserting that “wars originate in the structure of the international political system” (Waltz 1988, 617–119, 627).4

As the last straw, some defensive realists may think that Van Evera’s Causes of War is the definitive theoretical attempt to link the security dilemma with war, because Van Evera (1999, 117) takes offense-defense theory (hereafter, ODT) as the “master theory” of the causes of war, and many defensive realists may have mistakenly taken ODT as part of the larger theory of the security dilemma because the two theories seem to be so closely interknitted in the literature (e.g. Jervis 1978; Glaser 1994–1995, 54, 60–67), or worse, taken the two theories as essentially equivalent (Glaser 1992, 498n1; 1997, 172; Glaser and Kaufmann 1998, 44).5

Yet, despite their seemingly close association in the literature, ODT’s two key components, the distinguishability of offense and defense and


the offense-defense balance, are merely regulators of the security dilemma and certainly is not equivalent to the security dilemma. According to the original BHJ formulation, modifiers of the security dilemma are distinct from the security dilemma itself, and they cannot give birth to the secu-rity dilemma itself (chapter 2 this volume). Van Evera (1999, 117n1) was thus right to explicitly deny that ODT is an inherent part of the security dilemma theory.

Moreover, even if we take ODT as an integral part of the security dilemma theory, Van Evera’s hypothesis that “war is more likely when con-quest is easy or is perceived as easy” cannot tell us whether the security dilemma has been a major cause of war in history. His theory merely states that “war is more likely when conquest is easy or is perceived as easy” or “war is more likely to break out when one or two states believe that conquest is easy,” but nothing more. Without an explicit assumption on whether states are offensive realist states or defensive realist states (i.e., states’ inten-tions), Van Evera’s theory cannot explain whether war occur intentionally because of malign aggressiveness or unintentionally because of the security dilemma. Consequently, Van Evera’s work does not support the thesis that the security dilemma has been a major cause of war in history.

To summarize, most proponents of the security dilemma/spiral model have merely maintained that the security dilemma can cause war, but few have said anything explicit on whether the security dilemma has been a major cause of war.

What Defensive Realists Have (Not) Argued Empirically

Ultimately, the thesis describing the security dilemma as a major cause of war in history must be substantiated by empirical evidence. So, have defen-sive realists provided much empirical evidence on this front? The answer is again “No.”

To claim that the vicious spiral of power competition derived from the security dilemma has been a major cause of wars in history, two interre-lated sets of evidence are required. First, one has to show that the security dilemma has been prevalent in international politics—otherwise, the secu-rity dilemma will be unlikely to be a major cause of war, because war has been prevalent in history. Second, one has to show that most or many of the wars in history were indeed caused by the security dilemma. This section examines whether defensive realists have provided the second set of evidence, leaving the first set of evidence for analysis in section 4 later.

To show that most or many of the wars in history were indeed caused by the security dilemma, one has to show (1) that a war was indeed caused by a security dilemma and then (2) that such wars were frequent.


To show that a war was indeed caused by a security dilemma, at the very least, one has to show that (1) neither side in the conflict harbors malign intentions against each other at the beginning; (2) both sides act and react to exacerbate the fear between them; (3) at least one side eventually ends up in believing that preemptive or preventive war is the only viable solution for its security (because it believes that the other side is inherently aggres-sive and is close to launching an attack);6 (4) war eventually results from this dynamics. So far, proponents of the thesis that the security dilemma has been a major cause of war have yet to present any robust evidence to support their case. Instead, they have merely asserted that the dynamics of the security dilemma has often caused wars among nations.

The boldest claim that the security dilemma has been a major cause of war in history came from Taliaferro: “The historical record abounds with cases of states that pursued security-driven expansion or preventive war.” He then listed ten cases of war and expansion as supporting evidence. The ten cases are as follows: the Peloponnesian War, the War of Spanish Succession, the Seven Years’ War, the French Revolutionary Wars, Russia’s mobilization in the July 1914 crisis, the 1962 Sino-India War, the 1965 and 1971 India-Pakistan Wars and the 1990 Kashmir Crisis, Japan’s expan-sion in the 1930s, and China’s intervention in the Korea War (Taliaferro 2000–2001, 145, 147–149).

Yet, of the ten cases that Taliaferro identif ied as security dilemma-induced wars, only China’s intervention in the Korea War can be reasonably argued to be a war largely caused by a security dilemma (Reiter 1995, 23–25).7 Let’s examine some of the remaining cases of Taliaferro briefly.

For a start, Japan’s expansion in the 1930s was not pursued by a security-seeking state in the defensive realism sense, but by an imperialist state (i.e., an offensive realist state). Imperial Japan saw its security depen-dent upon annexation of Korea, Manchuria, and eventually the whole East Asia. As such, just as there was no security dilemma between Hitler’s Germany and other states, there also was no security dilemma between imperial Japan and any other states. Likewise, the three wars between India and Pakistan were not wars driven by the security dilemma either. The 1965 war and the 1990 Kashmir crisis between India and Pakistan were initiated by a Pakistan that was determined to reclaim Kashmir and Jammu, and the 1971 war was launched by an India that was determined to dismember Pakistan. Even the 1962 India-China war is a dubious case. While India and China might not have wanted war; India was intentionally threatening China’s (real or perceived) territorial integrity (Maxwell 1970; Whiting 1975). Finally, listing the French Revolutionary Wars as wars driven by the security dilemma is preposterous. Revolutionary France


could not possibly have been a defensive realist state; it was a revolutionary/revisionist or offensive realist power.

Thus, almost all the wars that Taliaferro identified as wars caused by the security dilemma were fought between two sides in which at least one side was intentionally threatening. Indeed, Taliaferro’s stand is severely undermined by his reliance on almost the same set of cases that Copeland (2000) employed to support the theory of dynamic differentials and war, an offensive realism theory that Taliaferro incorrectly classified as a defensive realism theory (see Tang 2008a and chapter 6 in this volume for details).8

Fundamentally, even if we admit Taliaferro to be correct in that those wars are “all cases where security-seeking state precipitate international crises and wars in part because of uncertainty of others’ intentions and impeding power shifts (Taliaferro 2000–2001, 147),” his statement cannot substantiate the thesis that the security dilemma has been the (principal) cause of those wars. This is so because security-seeking, uncertainty of others’ intentions, and (the concern about) impending power shifts do not make a situation a genuine security dilemma: A genuine security dilemma requires the lack of malign intentions from both sides to operate (chapter 2). Thus, even if Taliaferro was correct that security-seeking motives were the major force behind those expansions and wars, it does not mean that those wars were largely driven by the security dilemma. To identify a situation as a security dilemma, one needs to show a lack of malign intentions from both sides, and this critical ingredient of a genuine security dilemma is completely missing from Taliaferro’s discussion.

Instead, by equating pursued security-driven expansion or preventive war with wars caused by the security dilemma, Taliaferro was implic-itly arguing that as long as a state’s behavior (including preventive war or security-driven expansion) is driven by security-seeking motive, it is driven by a security dilemma. By doing so, Taliaferro is replacing lack of malign intentions with security-seeking motive, or taking the two as essentially equivalent. Yet, as shown in chapter 2, security-seeking motive is not equivalent to lack of malign intentions because all states (including offensive realist states) seek security under anarchy. Thus, to replace lack of malign intentions with security-seeking motive as one of the essential aspects of a genuine security dilemma is to violate the BHJ formulation of the security dilemma.

Hence, once we apply the more rigorous definition of the security dilemma to cases identified by Taliaferro, his claim that many conflicts have been caused by the security dilemma falls apart. Few of the wars identified by him were actually caused by the security dilemma per se; rather, those wars were mostly caused by predatory or offensive realist states with malign intentions.


Again, some defensive realists may claim that Van Evera (1999) has provided the definitive empirical support for the thesis that the security dilemma has been a major cause of war in history. In reality, however, even if we fully agree with Van Evera’s theory, his theory provides little empiri-cal support for the thesis that the security dilemma has been a major cause of war because it relies on a very limited number of cases for empirical support.9

Indeed, for defensive realists, the sole major historical case that has been argued to be a mostly security dilemma-induced or inadver-tent war has been WWI. But even this case remains a contentious case (Levy 1990–1991; Trachtenberg 1990). For instance, Kydd (1997b, 154) noted, “the four continental powers in WWI all had serious non-security related quarrels that played an indispensable role in producing the war.” As such, it is hard to argue that major warring states (especially Russia, France, Britain, Germany, and Austria) in WWI were defensive realist states. Likewise, Jack Snyder admitted: “it becomes tortuous to explain German behavior in terms of a security dilemma” (Snyder 1985, 174). In the end, perhaps the strongest claim one can make is that a real spiral had played a critical role in causing WWI, but real irreconcilable conflict of interest rather than a security dilemma was the more critical driving force of the war. More importantly, the criticalness of WWI for the thesis that the security dilemma has been a major cause of war actually indicates the rarity of such wars. Counterintuitively, if there have been many secu-rity dilemma-caused wars, why would defensive realists defend the case of WWI so earnestly?10

The Security Dilemma as a Major Cause of War: The Cons

With the thesis that the security dilemma has been a major cause of war in history resting on such shaky theoretical and empirical ground, it is not surprising that it has been repeatedly changed.

For instance, although his understanding of the security dilemma was mostly imprecise or incorrect (chapter 2), Jack Snyder concluded that the incentives for conflict derived from the “structural” (thus genuine) security dilemma “are rarely, if ever, this strong [to cause actual conflicts]” (1985, 157; emphasis added). Rather, according to Snyder, other factors such as misperceptions and/or offensive intentions from one or both sides must be present to drive states into war.

More recently, Schweller (1996) and Kydd (1997b) advanced the case against the thesis that the security dilemma has been a major cause of war in history more explicitly and systematically.11 Contending that defensive realist states (i.e., sheep, or security-seekers) rarely fight each other and it


has been offensive realist states (i.e., predators, wolves) that have caused most of the wars in history, Schweller and Kydd explicitly reject the thesis that the security dilemma has been a major cause of war.

Despite reaching a valid conclusion, however, neither Schweller nor Kydd (and Snyder, for that matter) provided a compelling explanation why the security dilemma has not been a major cause of war in history. Moreover, their arguments actually rest on some serious misunder-standings of the security dilemma (see chapter 2), and are thus subject to some valid counterattacks from defensive realists (e.g., Taliaferro 2000–2001).

This section critically examines Schweller and Kydd’s arguments to pave the way for a more definitive explanation why the security dilemma has not been a major cause of war in history.

Schweller (1996)

Schweller was certainly right to assert that the existence of the security dilemma is conditional. There is no security dilemma if one or both sides are aggressive (Schweller 1996, 117–119). Unfortunately, Schweller made a fundamental error at the very beginning of his elaboration: he pretty much eliminated the uncertainty about others’ present intentions—but especially future ones—under anarchy. He writes, “The security dilemma vanishes in a system composed of all status-quo states . . . [or] in a hypo-thetical world in which states are known to be status quo and cannot be otherwise in the future . . . [or] in a world without aggressors, in which all states are happy with what they currently have and are expected to remain so in the future” (118–119).12 Schweller got the condition of the security dilemma only half right.

To build his case that there is no real security dilemma in a world of all status quo states (i.e., defensive realist states), Schweller advanced two arguments. First, Schweller charged that defensive realists assume that all states are defensive realist states, taking Waltz (1979) as the prime target. Second, Schweller reasoned that if the world is made of only defensive realist states, states will be able to correctly perceive each other’s (benign) intentions. As a result, there will be very little uncer-tainty and the security dilemma will rarely exist in such a world. Both arguments are incorrect.

First, contra Schweller, Waltz did allow the possibility of greedy or predatory states because Waltz explicitly argued that states can desire anything between mere survival and universal domination (Waltz 1979, 118, 126). Likewise, Jervis, Glaser, and Snyder all have explicitly recog-nized the possibility of opportunistic expansionist, inherently greedy states, and Hitler (Glaser 1992; Jervis 1982, 616; 1999, 49; Snyder 1984,


462n2). Thus, even in the world depicted by these defensive realists, there is still uncertainty about others’ present and future intentions.

Second and more importantly, contra Schweller, even in a world of all defensive realist states, the security dilemma can still operate, because uncertainty about states’ present and future intentions is inherent under anarchy.

To begin with, even in a world of all defensive realist states, it is hard to imagine that states can know each other’s present intentions perfectly well. More critically, even if states can know each other’s present intentions perfectly well, they cannot possibly know each other’s future intentions perfectly well, simply because intentions can change. If so, then as long as anarchy reigns and there is no central authority to make states stick to their original (benign) intentions, this uncertainty about others’ present and future intentions—or the possibility that there may be greedy states out there now and in the future—cannot be completely eliminated. As such, even in a world of all defensive realist states, the security dilemma can still operate.

Fundamentally, Schweller ignored the critical fact that the uncertainty about others’ present and future intentions is indispensable for the secu-rity dilemma: “Without this uncertainty, the security dilemma is no dilemma” (Kydd 1997b, 126; emphasis added; see also chapter 2 this volume). Because Schweller assumed that in a world of defensive realist states, states can know each other’s present and future intentions perfectly well, he eliminated the inherent uncertainty under anarchy at the very beginning of his discussion (Glaser 1997, 189–193, 195; Kydd 1977a, 371, 383; Taliaferro 2000–2001, 144–145). Not surprisingly, Schweller was then able to claim that the secu-rity dilemma cannot exist in a defensive realism world.

Moreover, Schweller also eliminated fear, perhaps because fear among states is partly a function of the uncertainty about others’ intentions (Tang 2008a). Because Schweller believed that the states in a world of all defen-sive realist states will be able to know each other’s present and future inten-tions perfectly, he saw no reason for states to fear each other. For him, fear between two security-seeking states is “misplaced” (Schweller 1996, 117). Yet, states’ fear about each other is “misplaced” only if they can surely know each other’s present and future intentions perfectly. Because uncertainty is inherent under anarchy, some fear is not “misplaced,” but mandatory.

Kydd (1997b)

Explicitly dedicated to deal with structural (defensive) realists’ counter-charge that Schweller underestimated the impact of uncertainty about states’ intentions on the ability of security-seeking states to cooperate (Spirtas 1996, 410–412), Kydd (1997b) contended that the security


dilemma cannot drive two genuine security-seekers (i.e., sheep or defensive realist states) into war because the two states possess various means to signal their benign intentions to each other, thus reducing uncertainty about each other’s present intentions and being able to avoid war.13 Kydd believed that the cause behind the stand that the security dilemma alone leads to tragic results is that “structural realism strongly overestimates the difficulty in assessing state motivations.” To correct this overestimation, Kydd (1997b, 128) then asserted that “information on the motivations of security-seeking states is so easy to come by that mistaken fears about motivations cannot plausibly explain any significant war, arms race or crisis this [i.e., the twentieth] century.”

Despite reaching a correct conclusion, Kydd’s elaboration also rests on shaky ground. As Taliaferro (2000–2001, 144–147) pointed out, Kydd overcorrected structural defensive realism’s overconfidence in assessing state’s intentions. Put simply, just because states do have the possible tools for signaling benign intentions and correctly assessing other states’ inten-tions, it does not mean that they will employ those tools consistently. Likewise, just because states can employ those tools under all circum-stances, it does not imply that they can interpret those signals correctly all the time. Indeed, Kydd acknowledged this possibility in his earlier work (e.g. Kydd 1997a, 395–396).

Overall, Kydd neglected four factors or possibilities that can prevent states from signaling benign intentions and/or correctly reading others’ benign intentions, even over a fairly long period of time.

First, as a rational choice theorist, Kydd paid attention only to uncer-tainty, but not fear, another crucial psychological ingredient for linking anarchy with the security dilemma (thus an important conditioner of state behavior). As a result, Kydd failed to appreciate the possibility that under certain circumstances, fear—which is not mistaken but rather mandatory under anarchy—may prevent defensive realist states from pursuing many of the policies for achieving security among them that Kydd identified.14 Indeed, because states tend to exaggerate the hostility of others, they may view the other side as intentionally threatening, even if the other side is a genuine security-seeker who is taking some precautionary measures. As a result, a defensive realist state may fail to recognize that it is in a security dilemma with another fellow defensive realist state, thus failing to see the need for taking those measures of signaling benign intentions and assuring other states, as Kydd articulated.15

Second, because there are structural and perceptual modifiers that reg-ulate the severity of the security dilemma (chapter 2), sometimes it may be too risky, thus difficult, for a state to take costly measures to signal benign intentions under certain circumstances, even if the state recognizes that it (and the other state) may be in a security-dilemma situation (but it


can never be sure of its interpretation!).16 Both Kydd and Schweller have acknowledged this point implicitly earlier in their discussion, but seemed to forget it conveniently as they moved on.17 Third, even if a state does take costly measures to signal their benign intentions, other states may not always interpret those signals correctly or be willing to interpret those signals as signaling genuine benign intentions, because the other states may fear that the state signalling its benign intentions is merely out to trap them.18 Fourth, for much of human history, states might have not known how to signal benign intentions.19 Apparently, a combination of the four possibilities can lead to a worsening of the security dilemma, thus eventu-ally leading to conflict and war.

Additionally, Kydd (1997b, 126) is also wrong to argue that “defensive realists pay more attention to uncertainty over motivations, but typically consider it an aggravating factor for the security dilemma, rather than an essential underlying prerequisite without which there would be no dilemma,” because Glaser (1994, 67–70) did so. As shown in chapter 2, despite being a defensive realist, Glaser actually has some incorrect or imprecise understandings about the security dilemma, as indicated by his misinformed expanding of the security dilemma (Glaser 1992), his refusal to admit that Schweller was right in asserting that there would be no secu-rity dilemma between an offensive realist state and other states (Glaser 1997), and most glaringly taking the security dilemma theory as synony-mous with ODT (Glaser and Kaufmann 1998, 44). In sharp contrast, all three original proponents of the security dilemma explicitly take uncer-tainty (over motives and intentions) to be an essential prerequisite for a security dilemma to operate.

Because Kydd’ thesis is subject to these counterattacks, he cannot convince many defensive realists to abandon the thesis that the secu-rity dilemma has been a major cause of war in history, despite having considerably weakened the thesis.

Like Schweller, Kydd too underestimated the difficulty of seeking security under anarchy. Indeed, their errors were remarkably similar. Schweller simply eliminated much uncertainty by insisting that defensive realism assumes that every state is a defensive realist state and will remain so, while Kydd believed that uncertainty over intentions can be easily overcome. Neither position can pass close scrutiny.

The Security Dilemma Has Not Been a Major Cause of War: An Empirical and Normative Explanation

There are only two channels through which the security dilemma can lead to war: preemptive war and preventive war. In other words, for the security dilemma to be a major cause of war in history, it must eventually induce


states to either preempt or prevent, or both. Building upon the discussion above and in the previous chapter, this section sets out to show that the security dilemma has not been a major cause of either preemptive war or preventive war, thus providing a more definitive explanation why the security dilemma has not been a major cause of war. Essentially, I show that even if the security dilemma among nations has been widespread, the security dilemma has not been a major cause of war among nations in history. The next section, building upon this section, then seeks to under-mine the notion that the security dilemma among nations has been wide-spread, thus undermining the thesis that the security dilemma has been a major cause of war in history even more fundamentally.

Although preemptive war and preventive war are two distinct concepts, there have been some differences on how to differentiate these two types of war.20 Here, I shall provide the working definitions of these two types of war. Preventive war is primarily driven by a state’s concern about the (real or perceived) impending adverse shift of relative power. Preemptive war is primarily driven by a state’s concern about the imminent attack by an adversary, although this general concern usually includes a concern for the loss of relative power if the adversary is allowed to attack first. In terms of time-horizon, the time-horizon of preventive war is longer, while that of preemptive war is shorter. Of course, preemptive motive can be part of a preventive war if one or two states are thinking of launching a preventive war.21

Preemptive Wars Have Been Rare, Period!

The original BHJ formulation explicitly predicts that the major mecha-nism through which the security dilemma drives states into war is through heightening the preemptive motive (Jervis 1978; Reiter 1995). Yet, few proponents of the security dilemma-preemptive war thesis have bothered to establish the thesis’s empirical validity. Worse, two studies have provided compelling evidence that preemptive wars have been rare, period!

Explicitly dedicated to test the prediction of the spiral model (not the security dilemma),22 Dan Reiter (1995) provided strong evidences that preemptive war has been rare in history. After examining all the sixty-seven wars from 1816 to 1980 from the Correlates of War (COW) data-base, Reiter found only three wars (i.e., WWI, China’s intervention in the Korean War, and the 1967 Israeli attack on Egypt) that one can reasonably argue to be mostly caused by the preemptive motive.

Using a slightly different and more relaxed definition of preemptive war, Van Evera was able to find only four wars as mostly caused by a motive to preempt (1999, 41n19). Thus, despite emphasizing that preemptive


motivation may be a “potent catalyst of war” and complaining that Reiter might have been too stringent in demanding a war to be identified as a pre-emptive war,23 Van Evera nonetheless admitted that “preemptive motives of this sort are not commonplace (Van Evera, 1999, 41).” Such a conclu-sion is only slightly supportive for the thesis that the security dilemma often causes preemptive war than Reiter’s blunter assertion that “preemp-tive wars almost never happen” (Reiter 1995, 6).

More importantly, among the six wars identified by either Van Evera or Reiter as preemptive wars, only the war that resulted from China’s intervention in Korea can be argued to be a war mostly caused by a secu-rity dilemma. All other wars were initiated because one or both sides had aggressive designs, and the security dilemma does not apply in those cases. For instance, Bismarck wanted to initiate the 1870 Franco-Prussian War so that he could finally unify Germany and crush France, while Israel launched the 1967 Six-day war to crush Arab countries’ apparent attempt to eliminate Israel. Finally, even WWI, the iconoclastic case of the spiral model, was almost surely not caused by a security dilemma (Snyder 1985, 174; Kydd 1997b, 154; Lieber 2007).

In sum, the security dilemma cannot possibly have been a major cause of war if preemptive war is the major channel through which it causes wars: preemptive wars have been rare, and even those few cases of preemp-tive war were rarely caused by the security dilemma.

Security-driven Expansions or Preventive Wars Were Prevalent but Few of Them Were Caused by the Security Dilemma

The original BHJ formulation explicitly predicts that the major channel through which the security dilemma causes war is by heightening states’ motive to preempt (Jervis 1978), but it has not been explicit on whether the security dilemma can also cause preventive war. With the channel of preemptive war being severely undermined by existing work, however, it seems that many proponents for the thesis that the security dilemma has been a major cause war have increasingly turned to the channel of preven-tive war, or security-driven expansion, as the major channel through which the security dilemma causes war.24

More notably, to salvage the thesis that the security dilemma has been a major cause of war, Taliaferro asserted that “the historical record abounds with cases of states that pursued security-driven expansion or preventive war (Taliaferro 2000–2001, 145; emphasis added),” and then implied that these preventive wars or security-driven expansions were caused by the security dilemma. Taliaferro’s logic came down to two propositions. First, even genuine defensive realist states often do pursue security-driven


expansion or preventive war under the dynamics of the security dilemma (because they feel insecure). Second, security-driven expansions caused by the vicious spiral of the security dilemma have been frequent or even rampant.

On the surface, the thesis that the security dilemma has been a major cause of war would receive a significant boost if the security dilemma can indeed cause states to launch preventive wars, because there seems to have been quite a number of preventive wars from ancient to modern times (Copeland 2000). Moreover, the motive to launch preventive wars seems to be quite common among decision-makers. Thucydides spoke about Sparta’s fear about Athens’ growing power (Thucydides 1954, 49). Kaiser and Hitler feared about Russia’s rapid industrialization and its impending implications for Germany (Copeland 2000). The United States had con-sistently considered and employed preventive wars against its weak neigh-bors in the nineteenth century when it built its continental empire. During the course of the cold war, some U.S. decision-makers considered preven-tive war (in the form of nuclear strikes) against the Soviet Union and then China, and did launch preventive wars against the Dominica Republic (Leffler 2005).25

Although there have been counterarguments that preventive wars have been rare historically or that states rarely jump into the “window of opportunity” (Wolfers 1952, 188; Lebow 1984; Jervis 2005, 11), I shall give the benefit of doubt to the security dilemma-preventive war thesis for now. I shall show that even if preventive wars have been common in history, it does not mean that the security dilemma has been a major cause of war in history, for two interrelated reasons. First, most preventive wars (or security-driven expansions) are not driven by the security dilemma. Second, genuine defensive realist states do not easily fall for preventive wars. This subsection elaborates the first reason, and the next subsection, the second.

Essentially, a state will launch a security-driven expansion or preventive war with two essential conditions. First, a state must have a long, or even very long, time-horizon (Lee 2002–2003). Because relative power is an essential ingredient of security, a state with a long time-horizon fears the potential impact upon its future security of a possible relative decline.26 Second, a state must be so fearful that it assumes other states (as its poten-tial opponents) to be inherently aggressive. In other words, a state assumes the worst over others’ intentions. Because war between the state and its inherently aggressive opponents is inevitable, the state in real or imagined relative decline reasons that it has to expand now. Because other states are growing more rapidly, expansion will be more difficult in the future. In contrast, a successful expansion now can halt one’s relative decline because


it will bring in the spoils and thus expand one’s relative power versus that of the potential opponents. The logic of preventive war crucially depends on this worst-case assumption over others’ intentions (Tang 2008a), or “the inevitability of war” (Ripsman and Levy 2007, 40).27

As soon as a state assumes the worst over others’ intentions and decides to resort to preventive war, however, it becomes an offensive realist state. And as soon as this change of intentions occurs, the security dilemma—if there was one—no longer operates. Yet, if one can prove that such a change of intentions has been a result of the security dilemma dynamics—that is, a defensive realist state becomes so fearful due to the spiral of a security dilemma that it changes itself into an offensive realist state, one can then claim that the preventive war was indeed caused by a security dilemma. Indeed, this is the only way that the security dilemma can cause preventive war.

Alternatively, an offensive realist state can launch a preventive war regardless of the nature of its opponents. When this is the case, then obviously such a scenario is not a security dilemma. Its “security-driven” expansion can be taken as a step toward universal empire, and such a means is a means toward security that is only recognized by offensive realism (Mearsheimer 2001, 34). Indeed, the doctrine of preventive war was asso-ciated with Nazi Germany and Imperial Japan, both archetypical offensive realist states (Hendrickson 2002, 1). The Bush doctrine, by emphasizing preventive war (not preemptive war), is also an offensive realism doctrine (Jervis 2003).28

In sum, this leaves us with a scenario in which a defensive realist state mistakenly believes that a fellow defensive realist state is inherently aggressive (i.e., war is inevitable) and then decides to launch a preventive war. This is the only possible situation in which a preventive war is mostly caused by a security dilemma.

By eliminating several scenarios of preventive war as possible scenarios in which the security dilemma can play a major role in causing war, I have considerably weakened the thesis that the security dilemma has been a major cause of war because it induces states to launch preventive wars.

Yet, I shall further contend that even this last remaining scenario in which the security dilemma can be the major cause of the preventive war will also be small. A genuine defensive realist state does not launch a preventive war against a suspected aggressor (that is actually a like-minded defensive realist state), unless under extreme circumstances, even if it believes that the mind of the possible aggressor cannot be changed or the possible aggressor is going to launch a war in the not too distant future. Put differently, a defensive realist state will be extremely reluctant to launch a preventive war against another state. This is the question that I now turn to.


Defensive Realist States Do Not easily Fall for Preventive War

In danger of doing injustice to Jervis, I traced the intellectual foundation behind the notion that defensive realist states can easily fall for preven-tive wars to Jervis’s problematic assertion that “the security dilemma is at its most vicious when commitments, strategy, or technology dictate that the only route to security lies through expansion. Status quo must then act like aggressors (Jervis 1978, 187).” Here, Jervis suggested that under certain circumstances a defensive realist state will be compelled to behave like an offensive realist state.29 As a result, some defensive realists (e.g., Taliferro 2000–2001) now believe that a defensive realist state may often launch preventive war, just as an offensive realist state is supposed to (Mearsheimer 2001, 32).

This argument is mistaken. Under most circumstances, a defensive realist state can behave differently from an offensive realist state without having to sacrifice its core security interest if it chooses to do so. In other words, only under some extreme circumstances will a defensive realist state deem behaving differently from an offensive realist state as simply too risky. Yet, even when a defensive realist state believes that it will have to behave similarly like an offensive realist states, it does not mean that the defensive realist state will easily launch a preventive war. As a result, preventive wars among genuine defensive realist states will be extremely rare.

Resorting to preventive war may be irrational and counterproductive for two operational reasons. First, the perceived window of opportunity that is necessary for powering the preventive war may be false. If so, then launching a preventive war may prove to be self-defeating. Second, launching a preventive (or preemptive) war without providing conclusive evidence that an attack by another state is imminent carries heavy diplo-matic cost. A state that does so will not be able to gain allies in the future, because other states will deem the state as not only fundamentally unreli-able as an ally but also inherently threatening (Lebow 1984; Reiter 1995, 25–28, 33).30

Then there is a more fundamental reason a defensive realist state does not easily launch a preventive war (even if it suspects that the other state is inherently aggressive and is going to launch a war in the not too distant future), unless under extreme circumstance. Launching a preventive war easily is simply inconsistent with the general operational code of defensive realist states (see chapter 4 this volume).

First, resorting to a preventive or preemptive war is like reacting to the worst-case scenario or possibility. Yet, according to defensive realism, acting to a worst-case scenario under a suspected security dilemma is dan-gerous and counterproductive (Glaser 1997, 197–198). In fact, only offen-sive realist states consistently plan and act according to the worst-case


scenario (Tang 2008a; see also Copeland 2001, 213–214; Mearsheimer 2001, 45).

Second, when a defensive realist state launches a so-called “security-driven” expansion and preventive/preemptive war against a possible fellow defensive realist state, it is dismissing cooperation and choosing compe-tition as the only viable means of self-help. Yet, defensive realist states usually do not do so. Recognizing that cooperation is a viable means of self-help, under many circumstances, genuine security-seeking states can try to avoid conflict and war by pursuing cooperation by sending costly signals (Glaser 1994–1995; Kydd 1997b; 2000a). Again, only offensive realism generally dismisses cooperation as a viable means of self-help under anarchy (Grieco 1988; Mearsheimer 2001, 51–53).

Third, preventive war is a move that seeks to escape from a security dilemma (if it is real), thus a move to make a possibly avoidable tragedy an inescapable reality.31 Such a move is again more consistent with offensive realism (Layne 2003, 123, 128). By arguing that security-seeking states can also pursue expansion when feeling insecure—they expand because they must—some defensive realists have been essentially arguing for the same measure to escape from, not to alleviate, the security dilemma (Glaser 1994–1995, 55; Jervis 1978, 187; Herz 1955; Taliaferro 2000–2001, 145). Such a stand fundamentally contradicts defensive realism’s stand that the security dilemma can only be alleviated but not escaped from.

Thus, the notion that defensive realist states can easily fall for preventive wars is not very tenable. In sum, security-driven expansion or preventive war is generally inconsistent with the operational code of defensive real-ism. A genuine defensive realist state will not pursue (security-driven) expansions under most circumstances. The operational code of a defen-sive realist state is to maintain deterrence and defense, and then seek to change the mentality of the (suspected) offensive realist state through peaceful means. Facing a suspected offensive realist state does not auto-matically compel a defensive realist state to expand (see chapter 4 this volume).

As a result, the situation in which the security dilemma between two genuine defensive realist states has driven the two states to preventive wars must have almost never happened (Kydd 1997b, 149, 152). Most expan-sions or preventive wars in history, security-driven or not, had been the handiwork of offensive realist states rather than that of defensive realist states (Kydd 1997a, 373). As Lebow (1984) noted, during the cold war years, almost all those in the United States who believed that the Soviet Union would launch a preventive war against the United States and thus also advocated a preventive war against the Soviet Union were known hawks or hardliners (i.e., offensive realists).


A perhaps more striking contrast was the events that led to the out-break of WWII. Whereas Britain and France, two states that were close to defensive realist states at that time, at least when it comes to the European continent, had good reasons (with some hindsight, of course) to launch a preventive war against Hitler’s Germany around 1936–1939; however, they refrained from doing so (Ripsman and Levy 2007). In contrast, Japan did launch a preventive war against the United States (Van Evera 1999, 89–94), and Hitler invaded the Soviet Union partially due to a preventive motivation (Copeland 2000a).32

Defensive realists are kidding themselves by contending that defen-sive realist states can easily fall for (security-driven or not) expansions and preventive (or preemptive) wars (e.g. Glaser 1997; Taliaferro 2000–2001, 145). Doing so simply blurs the line between defensive realism and offen-sive realism and the line between an offensive realist state and a defensive realist state, thus allowing offensive realism to hijack some of the core arguments of defensive realism.33

It is possible that a defensive realist state will consider preemptive strike as an option when it is almost absolutely sure that (1) it is facing an inherently greedy and aggressive (offensive realist) state; (2) the of-fensive realist state is preparing to attack;34 (3) the first-strike advantage is potentially enormous. Because it will be rare that these circumstances coincide,35 preemptive wars launched by defensive realist states against of-fensive realist states will be extremely rare. In other words, compared to offensive realist states, which consistently think about and often launch preventive wars, defensive realist states have much more stringent condi-tions for launching preemptive wars. As such, it must have been extremely rare in history that preemptive wars were launched by defensive realist states. After all, in perhaps one of the strongest cases of the likelihood of preemptive war, Britain and France did not seriously pursue the option of a preemptive war against Hitler’s Germany even though they were officially at war against each other after 1939.

Summary

In this section, I have provided a normative argument about why the security dilemma could not have been a major cause of war in history, com-bining existing empirical evidences and theoretical works. Having denied the security dilemma a prominent place in the making of war in history, my argument immediately calls for a more fundamental question: If the secu-rity dilemma has not been a major cause of war in history, does it have any value in the studies of international politics at all? And to put the question in an even larger context: Without the security dilemma as a possible cause


of war, does defensive realism have a legitimate claim as a grand theory of international politics? These are the questions I now turn to.

Why the Security Dilemma Has Not Been a Major Cause of War: A Social Evolutionary Explanation

After advancing an empirical and normative explanation why the security dilemma has not been a major cause of war in history, this section advances a social evolutionary explanation regarding why this was so.36 The thrust of the explanation can be summarized as: the security dilemma is historical, not ahistorical.

Partly because a genuine security dilemma must be structural in origin (chapter 2), many hold the impression that the structure of international politics (i.e., anarchy) inevitably leads to the security dilemma (Taliaferro 2001, 131, 136n24; see also Glaser 1997; Wheeler and Booth 1992, 30, 43). In other words, many believe that the security dilemma is an inherent property of anarchy, thus ahistorical. This impression, however, is simply false: The security dilemma is not an inherent property of anarchy; rather, it is a historical construct in international politics (Schweller 1996, 91, 117–119; Wendt 1992, 401–402, 407; 1995, 73, 77).

Let us now recall that the original BHJ formulation explicitly dictates that the existence of the security dilemma is conditional, not universal (chapter 2). More specifically, the security dilemma operates only between two genuine defensive realist states, but not between two offensive realist states or between an offensive realist state and a defensive realist state.

This conditionality of the security dilemma necessarily means that the nature of the majority of the states in the international system matters critically for the degree of prevalence of the security dilemma. Straightforwardly, if the majority of the states in the system are offensive realist states, there would be very few security dilemmas operating in the system. Only if the majority of the states in the system are defensive realist states, would the security dilemma be prevalent within the system.

Thus, contrary to what Taliaferro (2000–2001, 136n24) asserted, the security dilemma will not be widespread in just any anarchic environment. In other words, the nature of a particular anarchy matters critically for the security dilemma to be prevalent in that particular anarchy.37 More spe-cifically, the security dilemma cannot possibly be prevalent in an offensive realism world in which most states intentionally threaten each other (either because of their nature or because of anarchy). In such a world, the logic is “to-kill-or-get-killed,” and there will be very few security dilemmas, only a lot of security or power competition (Snyder 2002, 156). Only in a defen-sive realism world in which most states no longer intentionally threaten


each other and the logic is “to-live-and-let-live” can the security dilemma be prevalent.

As I argue elsewhere, the world had been an offensive realism world for most of our history, and the defensive realism world that we enjoy today came into existence only very recently (Tang 2009b).38 As such, for much of our history, most states had been offensive realist states and they lived in an offensive realism world. In such a world, there were really very few security dilemmas.

Indeed, for most of our history, to-expand-or-to-be-expended (by expansion) was the rule, and there were a lot of conquests and expansions going on: “Throughout history a principal objective of states has been the conquest of territory” (Gilpin 1981, 23). Despite—or more fittingly, precisely because—the security dilemma was mostly absent, war had been rampant in human history: those conquests and expansions were not merely causes of war, they were the actual wars.

Thus, the security dilemma has not been a major cause of war in history, even if we want to call most wars in history as “security-seeking” wars (because all states seek security under anarchy), simply because the prominence of the security dilemma in international politics itself has been a product of the evolution of international politics and it has been a very recent phenomenon.39 Overall, once we admit that the existence of the security dilemma is conditional and thus historical, the thesis that the security dilemma has been a major cause of war in history becomes utterly untenable, no matter how vicious the spiral of power competition derived from the security dilemma can be. Defensive realists have been wrong in implying or insisting that the security dilemma has been a major driver of conflicts and war among nations for most of, if not the whole of, human history. Instead, predatory or offensive realist states that seek conquests and expansions rather than security dilemmas (among defensive realist states) have been the major perpetrator of wars among nations in history (Kydd 1997a, 373; 1997b, 149–152; Schweller, 1996; Diamond 1997; Tang 2009b).

Causes of the Confusion

This section highlights the three major reasons that cause many to hold the view that the security dilemma has been a major cause of war in history. By doing so, I reinforce the points made in the previous chapter that there have been many areas of confusion in our understanding of the security dilemma, and that a more rigorous definition and application of the concept is necessary if defensive realism is to anchor much of its claim as a grand theory of international politics upon the security dilemma.


Conflating the Spiral with the Security Dilemma

The first major reason many believe that the security dilemma has been a major cause of war in history is that they have a loose definition of the security dilemma. Consequently, they tend to conflate a spiral with a security dilemma. Because the spiral is universal and a security dilemma can and often do lead to a spiral, many simply take all spirals as security dilemmas. And because most wars in history were preceded by spirals (i.e., heightened tension), many then come to believe that most wars were preceded or even caused by a security dilemma.

A security dilemma, however, is different from a spiral. The security dilemma applies to only a subset of the spirals in which neither side harbors malign intentions (see chapter 2). Thus, just because there was a spiral (e.g., an arms race) that precedes a war does not mean that the security dilemma was the cause, or even the major cause, of the war. In fact, many, if not most, spirals are not security dilemmas or caused by security dilemmas, but are rather the work of offensive realist states or driven by intentionally malign behaviors (Jervis 1999, 49; Kydd 1997a, 373). As such, whereas the spiral has fueled many conflicts, the security dilemma has not.

For instance, the race in naval expansion between Britain and Germany before WWI certainly played an important role in driving the two states into war, and this naval race certainly resembles a classic spiral. Yet, the naval race, despite being a genuine spiral, was not a spiral driven purely or even largely by a security dilemma. On the one hand, Britain, as an imperial hegemon, wanted to maintain its naval supremacy to protect its far-flung empire (and to maintain an empire is a typical behavior of an offensive realist state). On the other hand, Germany wanted to catch up with England because it wanted “a place under the Sun.” For a powerful posi-tion, it was willing to accept high risk in order to challenge Britain. Thus, both Britain and Germany before WWI were offensive realist states, and there was no real security dilemma between them. While the two offensive realist states were in a spiral of naval races, the spiral was driven by their expansionist strategic interests, not by a security dilemma. Consequently, the war between the two could not possibly have been mainly driven by a genuine security dilemma, despite the naval race between them being a real spiral.40

Likewise, many may believe that the much quoted (thus frequently abused) passage of Thucydides—that “the growth of the power of Athens, and the alarm which this inspired in Lacedaemon, made war inevitable”—was about security dilemma. But this interpretation vastly overstretches the concept. Both Athens and Sparta were imperial powers, and it is difficult to imagine that there was a security dilemma between the two. Thucydides


certainly singled out power shift and fear as two major causes of the war, but he did not necessarily speak about the security dilemma. Fear and power alone do not make a situation a genuine security dilemma.

The Tied Hands of Defensive Realists

The second reason many (defensive) realists believe that the security dilemma has been a major cause of war is purely instrumental. As a grand theory of international politics, defensive realism seeks to explain the prevalence of war in history. At the same time, however, defensive realism also implicitly, if not explicitly, assumes that states mostly (should and do) seek security with a defensive approach (i.e., few defensive realist states will intentionally seek conquest). If so, an obvious challenge arises: how can defensive realism explain the recurrence of war among states in history if most states are assumed to be defensive realist states?

It is here that the security dilemma comes to defensive realism’s rescue. Between two defensive realist states, the vicious dynamics of the security dilemma seems to be a neat, and perhaps the only plausible, mechanism that can cause war between them: “Without it [i.e., the security dilemma] a system of rational states motivated solely by security would never generate security competition, arms races or war” (Glaser 2003, 406; emphasis added). Because most, if not all, defensive realists believe that defensive realism will have no legitimate claim to be a grand theory of international politics unless it can explain the prevalence of war in history well enough, defensive realists are somewhat forced to argue that the security dilemma has been a major cause of war in history.41 This leads defensive realists to explicitly or implicitly contend that the security dilemma has been preva-lent in history and that defensive realist states will easily fall for preemptive and/or preventive war when they are trapped in a vicious security dilemma. Then, to maintain consistency between their “status quo-biased” theory and the prevalence of war in history, some defensive realists have, perhaps unconsciously, expanded the concept of security dilemma to cover almost all situations. Together, defensive realists ended up vastly exaggerating the possible role of the security dilemma in causing war. Such a stand, however, is incorrect and unnecessary.

An Un-evolutionary Approach toward International Politics

Many defensive realists seek to explain the bulk of, if not the whole of, human history.42 Because war seems to have been a pervasive feature of human society in history and this spiral mechanism of the security dilemma seems to be the only mechanism through which two defensive


realist states can end up fighting with each other, defensive realists were forced to argue that the security dilemma must have been a major cause of war in human history. The possibility that the security dilemma is impor-tant for understanding and managing international politics today (and for the foreseeable future), even if it has not been a major cause of war in history, has never been seriously considered, because for most defensive realists, admitting this possibility will fundamentally nullify defensive realism’s claim for being a timeless theory of international politics.

This imagined necessity to claim that the security dilemma has been a major cause of war and consequently the practice to twist it is based on an un-evolutionary approach toward social changes. An approach claiming that only if something had been important in history could it be important for our present and future. If something was not important in history, it cannot possibly be important for the present and future.

Yet, once we adopt a social evolutionary approach toward international politics, it becomes clear that just because the security dilemma has not been a major cause of war does not necessarily mean that it cannot be the linchpin concept of defensive realism as a grand theory of interna-tional politics. In fact, a social evolutionary approach toward international politics points to the very possibility that the security dilemma may have become and may well remain important for the more recent history and today, precisely because the security dilemma has not been important for much of our earlier history. Because the human society has always been an evolutionary system, what was important yesterday can be unimportant today and what was unimportant yesterday can be important today.

More specifically, our world has indeed evolved itself from an offensive realism world of yesterday, in which most states were offensive realist states and the security dilemma was rare, to a defensive realism world of today, in which most states are defensive realist states and the security dilemma is prevalent (Tang 2009b). As such, precisely because the security dilemma has not been a major cause of war in earlier years, it becomes important for understanding it in international politics today.

Concluding Remarks

Just because the theory of security dilemma points to a possibility of “security-driven” expansion and preventive war, it does not mean that most or even many “security-driven” expansions or preventive wars were caused by the security dilemma. The security dilemma has not been a major cause of war in human history, and it has been offensive realist states (i.e., greedy states, predators, or wolves) that have been the major perpetra-tors of war for most of our history. Thus, while it is wrong to deny that


the security dilemma cannot potentially lead to war, it is equally wrong to believe that many, if not most, wars in human history were caused by the security dilemma.

Defensive realism does not have to inflate the importance of security dilemma in causing war and marginalize the importance of other obvious causes of war—that most wars in history were actually perpetrated by offensive realist states—in order to warn states of the danger that defen-sive realist states may end up fighting with each other. From a social evolutionary perspective, the security dilemma can, and should, be the linchpin concept of defensive realism as a grand theory of international politics, even though it has not been a major cause of war in history. In fact, a social evolutionary approach toward international politics points to the conclusion that precisely because the security dilemma had not been a major cause of war in the past, it is now important for understanding and managing international politics today.

Our world has firmly evolved from an offensive realism world in which almost every state was a predator into a defensive realism world of today in which most states are sheep. Whereas the security dilemma was rare and thus insignificant for causing war in the past, it has now become prevalent and thus important for understanding today’s world. Thus, the possibility that the vicious dynamics of security dilemma can drive two genuine defensive realist states into war now becomes too important to be discounted for understanding and managing international politics. In the defensive realism world of our time, in which predatory states are few and most states are defensive realist states (Tang 2009b), security dilemma becomes the more likely cause of war among states. As a result, educating statesmen about the danger of letting the security dilemma drive them into war becomes a major task for defensive realists for the sake of maintaining peace and avoiding conflicts in the present and future.

Ironically, policymakers realize the danger of inadvertent war driven by the vicious spiral of the security dilemma, and showing them the way out of this situation may well reduce the likelihood of such wars.43 Hence, defen-sive realism’s educational exercise will actually disprove one of defensive realism’s key hypotheses, thus becoming a self-denying prophecy (Jervis 1976, 81; 1997). Considering the destructiveness of war, however, this may be a small scientific price to pay.

Part II

Str ategic Choices


Chapter Four

The Oper ational Code of Defensive Realism

The preceding chapters clarify several critical areas of confusion that have come to obscure our understanding of the critical differences between defensive realism and offensive realism, redefine the security dilemma as the linchpin concept of defensive realism more rigorously, and resolve the critical question of security dilemma and war. With these clarifications, we can now better understand the differences between defensive realism and offensive realism in the two principal issue areas in international politics: war and peace, conflict and cooperation. By understanding the impor-tant differences (and commonalities) between the two realisms, we can now advance an operational code of defensive realism.1 Such an explicitly articulated operational code should greatly facilitate testing some of the theoretical claims of defensive realism versus other grand theories. More importantly, this general “operational code” should guide states’ security strategies today, because the international system has firmly evolved into a defensive realism world (Tang 2009b).

By explicitly contrasting the operational code of defensive realism with that of offensive realism, I also make many specific though yet unarticu-lated aspects of offensive realism more explicit. As such, the chapter also contributes to the articulation of a more detailed operational code of offen-sive realism. It shall become clear that the operational code of offensive realism is fairly simple and straightforward: to ensure that other states do not and cannot threaten oneself, and to constantly seek and take advantage of opportunities to weaken all other states, pausing only when temporary alliance is necessary (Mearsheimer 1994, 11–12; 2001, Chapter 2; see also Taliaferro 2000, 158–159).2

The rest of the chapter is divided into four sections. Section 1 outlines a ladder of strategies from a realism perspective. Section 2 examines the two realisms’ differences on war and peace. Section 3 examines the two realisms’ differences on conflict and cooperation. Section 4 outlines the


general operational code of defensive realism. A concluding section sum-marizes the differences between offensive realism and defensive realism when it comes to making strategies.

Realism’s Ladder of Strategies

In order to formulate an operational code of defensive realism, we need to have a general idea about the two realisms’ general preferences for strate-gies in facing another state. To do so, we have to first formulate a ladder of strategies for realist states. This section takes on this task by building upon two earlier attempts.

Two Earlier Attempts

Schweller (1998, 59–91; 1999, 7–18) and Copeland (2000, 38–41) devel-oped two different but somewhat related ladders of strategies through a realism prism.3 For a state (in relative decline) that is facing a rising state, Schweller listed six options: preventive war, balancing, bandwagoning, binding, engagement, and distancing/buckpassing. For the same situation, Copeland listed five options (from the least to the most confrontational): reassurance, doing nothing, deterrence/containment, crisis initiation, and preventive war. In addition to being incomplete, both schemes suffer from serious conceptual deficiencies.

Schweller’s scheme mixes strategies for dyadic situations with alliance strategies. Schweller failed to grasp that a state’s alliance strategy is usually part of its grand strategy. As a result, not only was he unable to provide a thorough discussion on either of the two sets of strategies, his elaboration also proved to be confusing. For instance, Schweller (1999, 17) treated Britain’s alliance with Japan between 1900 and 1907 when both faced a Russian threat in East Asia as a strategy of engagement. Earlier, how-ever, he also identified this British move as a strategy of “binding” (13). Yet, Britain’s “engagement” with Japan is better understood as a form of external balancing through alliance when facing a common threat (i.e., Russia), and binding is simply a natural outcome of allying with another state (Snyder 1984).

Copeland’s scheme fares better because he focused on strategies for dyadic situations and did not bring alliance strategies into the picture. His scheme, however, still has several major problems, the chief being that he did not define any one of the strategies in his scheme in any detail, and his understanding of the conditions for different strategies is problematic, to say the least.

Critically building upon the two schemes, I reformulate a more com-plete and rigorously defined ladder of strategies for realist states. First, I focus

operational code of defensive realism / 101

exclusively on a dyadic situation. By doing so, I am able to exclude alliance strategies from the discussion of grand strategies, although I do offer a ladder of alliance strategies at the end of this section and stress that a state’s alliance strategy is generally an integral part of its grand strategy. Second, I redefine different strategies more rigorously and clarify some areas of confusion. Third, I eliminate the (offensive realism) bias of the two schemes by including strate-gies that are consistent with both offensive realism and defensive realism.

Two caveats are in order. First, each strategy is defined as a pure strategy, yet states may pursue a mixture of several strategies, or at least a mixture of different strategic moves. Second, in the real world, each (pure) strategy contains a series of strategic moves. When discussing states’ actual behav-ior, a strategy must be differentiated from a particular strategic move.

A More Rigorous Ladder of Strategies

A strategy of appeasement is to retreat repeatedly (i.e., being accommo-dating or conciliatory) despite the adversary repeatedly taking advantage of one’s goodwill and pressing with its aggressive goals (Edelstein 2002, 5; Resnick 2001, 561–562; Ziring et al. 1995, 264–265).4 Defined as such, a conciliatory gesture or concession can be regarded as an appeasement ges-ture only if the other side is aggressive. In contrast, a conciliatory gesture or a concession when the other side is not aggressive should not be conflated with an appeasement gesture, even if the gesture was not reciprocated ini-tially (see chapter 5 for details).

A reassurance attempt is essentially an invitation toward genuine coop-eration to the other side, in the hope of moving toward more extensive cooperation and moving the relationship toward a more cooperative mode. A reassurance strategy, which consists of many reassurance attempts, can-not be an independent strategy by itself: It is usually a part of a strategy called “engagement” (see chapter 5).

Contra Schweller and Wohlforth that “engagement is simply a new, ‘more acceptable’ term for an old policy that used to be called appease-ment” (Schweller and Wohlforth 2000, 81; see also Mearsheimer 2001, 163–164; Schweller 1998, 74; 1999, 14; Craig and George 1995, 156–157, and Johnston and Ross 1999, xii, xiv–xv), engagement and appeasement are distinctly different (Resnick 2001, esp. 554–559; see also Edelstein 2002, 5). By equating engagement with appeasement, Schweller and Wohlforth basically reject any strategy that explores the possibility of cooperation as “appeasing,” thus debilitating and dangerous. This competitive bias is con-sistent with offensive realism, but not defensive realism.

Yet, as Schweller stated explicitly earlier, “Engagement is more than appeasement . . . engagement is most likely to succeed when the established


powers are strong enough to mix concessions with credible threats, to use sticks as well as carrots, in their attempts to satisfy the rising power” (Schweller 1999, 14–15; emphasis added). More precisely, the strategy of “engage-ment” contains three major components. First, it seeks to reassure the other side that one is not threatening. Second, it extends an invitation to coop-erate (with a possibility toward extensive cooperation) in order to gauge the other side’s intentions (see chapter 5 for a more detailed discussion). Third, it hedges against the possibility that the other side is an incorrigible aggressor. When the other side keeps pressing ahead, a state that pursues the strategy of engagement does not retreat repeatedly, because it has pre-pared a more hardened policy.5 Engagement thus has both a reassurance element and a defense/deterrence element embedded inside. At the core of engagement, there is an inherent hedging element and a difficult trade-off between reassurance and deterrence/defense within the strategy (Wolfers 1952, 497; Stein 1991, 451).

Thus, engagement, when properly understood and properly crafted, can serve four critical purposes simultaneously: (1) to assure the other side of one’s own benign intentions and to gauge the other side’s true intentions without putting one’s own vital interest at risk; (2) to deter the other side or, at least, to hedge against the possibility that the other side is actively thinking of aggression;6 (3) to change the other side’s intentions if the other side is not an incorrigible aggressor even if it may be malignant at the beginning; (4) to buy time to arm against the possibility that the other side is really an incorrigible aggressor (Schweller 1999, 14–16). Moreover, the second and third purposes can only be served by an engagement strategy (see chapter 5).7

Fourth, the strategy of containment can, and should, be further dif-ferentiated into two different forms: passive and active. A purely passive containment approach is a combination of deterrence and defense, without either an element of reassurance or an element of active rollback. A passive containment strategy reacts to other sides’ aggressive moves but does not actively provoke the other side.

In contrast, a state that pursues a strategy of active containment does not merely defend against and deter another state. It also actively initi-ates confrontations either to gain advantages or seek excuses for overt aggression. Active containment is thus a more hardened strategy than passive containment.8 Understood as such, Copeland’s “crisis initiation” (Copeland 2001), or provoking more generally, is an integral part of an active containment strategy.

Finally, and most critically, the option of security cooperation, which has been completely missing in Schweller’s and Copeland’s schemes, is cen-tral to defensive realism’s approach toward security (Glaser 1994–1995;


Jervis 1978).9 Here, it is necessary to bring another fine differentiation into the discussion: seeking extensive cooperation with another state and extending an invitation to cooperate to another state are two very different things. Defensive realism recognizes that seeking extensive cooperation among states is conditional. Specifically, defensive realism recognizes that seeking extensive cooperation is a viable means of self-help for a defensive realist state only when it faces a like-minded defensive realist state (Jervis 1982, 361; 1999, 50; Glaser 1994–1995). In contrast, seeking extensive cooperation is likely to be disastrous or even suicidal for a defensive real-ist state when facing an offensive realist state (i.e., it is in a false secu-rity dilemma) because the defensive realist state may end up being the ultimate sucker.

Defensive realism, however, also recognizes that when facing a state with unknown intentions, a defensive realist state will still be well-advised to explore the possibility of cooperation but at the same time guard against the possibility that the other state may be an offensive realist state. And it is under this circumstance that reassurance, that is, extending an invitation to cooperate, comes into the picture.10

Defensive realism understands that whereas pursuing extensive coop-eration is a viable strategy only when a defensive realist state faces a like-minded defensive realist state, extending an invitation to cooperate is a viable strategy under most circumstances if a defensive realist state takes some precautionary measures to guard against the possibility that the other side may be an offensive realist state. Contrary to what many (offen-sive realists) have asserted or believed, to learn another state’s true inten-tions by extending an invitation to cooperate, when it is done properly, does not have to be such a gamble that it risks a state’s survival: to explore the possibility of cooperation “need[s] not to be so crippling” (Snyder 1985, 179).11

Overall, defensive realism recognizes that extending an invitation to cooperate is an effective means of gauging other states’ intentions, and the condition for its applicability is far less demanding than the condition of pursuing extensive cooperation. Also, whereas seeking extensive cooper-ation can be an independent strategy when facing a like-minded defen-sive realist state, extending an invitation of cooperation to a state with unknown intentions functions as a part, although an indispensable part, of an engagement strategy.

With these clarifications, realism’s ladder of (grand) strategies, from the least confrontational to the most confrontational, runs from appease-ment, to doing nothing, to extensive security cooperation, to engagement, to passive containment, to active containment, and, finally, to preventive war (figure 4.1).

Figure 4.1 Realism’s ladder of strategies.

Preventive war

Active containment:provoking and roll-back

Passive containment: defense anddeterrence without reassurance

Engagement: defense and deterrencewith both a reassurance and a

hedging element

Def

ensi

ve r

ealis

m’s

pr

efer

ence

s

Offe

nsiv

e re

alis

m’s

pref

eren

ce

Do-nothing

Most confrontational: hard-lineA

berr

ant b

ehav

ior:

gen

eral

lyin

cons

iste

nt w

ith r

ealis

m u

nles

sun

der

extr

eme

circ

umst

ance

s

Security cooperation: arms control,arms reduction, and beyond

(i.e., CBMs)

Appeasement

Least confrontational: soft-line


Finally, a few words on alliance strategy are in order. When a state faces another state in a system with more than two of them, the ladder of alli-ance strategies for the first state toward the second state, from the friend-liest to the least friendly strategy, runs from allying with the second state, to explicit neutrality (distancing), to buckpassing/hiding, and to allying against the second state (figure 4.2).12

Alliance strategy, however, is not an independent strategy by itself under most circumstances. An alliance strategy is usually a part of a grand strategy. For instance, alliance strategies such as staying neutral or buck-passing/hiding should be considered as part of a grand strategy of “do-nothing” against the potential threat. Meanwhile, when a state faces two potential aggressors, and one of the two aggressors is less greedy, the state may be rational to appease the less greedy aggressor so that the state can concentrate resources in balancing the greedier aggressor or even gain a valuable ally—however temporary the alliance may be—against the greedier aggressor (Mearsheimer 2001, 164–165; Triesman 2004). Here,

Friendliest

Least friendly

Buckpassing/hiding

Allying with the state

Explicit neutrality, distancing,and hedging against both sides

Allying against the state

Figure 4.2 Realism’s ladder of alliance strategies.


alliance via appeasement is not an independent strategy but part of a strat-egy of balancing against the greedier aggressor.

Defensive Realism on War and Peace

With the discussion above, we can now state defensive realism’s positions on war and peace more rigorously. I do so by contrasting defensive real-ism’s positions with those of offensive realism.

War and Peace: General

Offensive RealismFor offensive realism, maximizing relative power constitutes the only via-ble means of self-help. As a result, states must constantly think of gaining power at the expense of other states, with offensive means (Mearsheimer 2001, 2, 32–36).

In an offensive realism world, every state is aggressive, thus a poten-tial opponent, whether by nature or because of anarchy. As such, states must jump on every possible opportunity of weakening each other (Elman 1996, 29; Labs 1997; Mearsheimer 2001, 29). In such a world, one state’s security must mean a decrease in another state’s security. Consequently, conflict of interest among states is not only genuine, but also genuinely irreconcilable. As such, war among states is inevitable, literally.

In an offensive realism world, war can only be ultimately avoided or prevented by becoming a universal empire so that there are no other states in the system, or, at the very least, by becoming a regional hegemon so that no other state is insane enough to challenge the hegemon. For offensive realism, there are very few conditions that are conducive to peace. Peace is relatively robust only under unipolarity (global or regional hegemony), balance-of-power (e.g., bipolarity or balanced multipolarity),13 geographi-cal barrier (e.g., large body of water), and robust (conventional or nuclear) deterrence (Copeland 2000; Gilpin 1981; Mearsheimer 1992, 2001; Organski and Kugler 1980; Wohlforth 1999; Waltz 1979).14

Thus, for offensive realism, peace is merely the intermission between wars, and it has no dynamics of its own. Peace is only possible when states are busily preparing for the next war, or, more precisely, when none of the states feel advantageous to start a war now. For offensive realism, there is no genuine peace as long as states have to compete for power (Mearsheimer 2001, 53).

For offensive realist states, politics is (almost) all about preparing for and winning the inevitable war. Offensive realism is thus ultra- Clausewitzian: War and politics are the same. In an offensive realism world, war among


states is intentional, thus not “tragic,” despite some offensive realists (e.g., Labs, Mearsheimer) attributing the ultimate cause of offensive behavior and war to anarchy.

Defensive RealismIn a defensive realism world, states generally seek security with defensive approaches and they do not constantly seek and grab opportunities to intentionally threaten others. In such a world, genuine conflicts of interest among states are still common, but they do not have to be genuinely irrec-oncilable all the time. Consequently, in a defensive realism world, war is not inevitable.

Defensive realism recognizes that it is extremely difficult, if not impos-sible, to avoid war between two offensive realist states. Defensive realism does not deny the possibility of war between two defensive realist states (due to the dynamics of the security dilemma). Defensive realism, how-ever, does not believe that war is inevitable for any combination of two states. Specifically, defensive realism believes that, most of the time, war between two defensive realist states represents a real and avoidable trag-edy. For defensive realism, even stable peace is possible between two like-minded defensive realist states if (and perhaps only if) both states take measures to signal their benign intentions, recognize each other’s benign intentions correctly, and pursue (reciprocal) cooperation persistently.

Between an offensive realist state and a defensive realist state, the pic-ture is more complex. The outcome depends first on whether the defensive realist state can maintain a robust deterrence and defense posture against the offensive realist state. Stable peace between these two states, however, depends on whether the defensive realist state can successfully convince the offensive realist state that aggression will not pay and becoming a like-minded defensive realist is a better way for its security, through influenc-ing the latter’s domestic politics. If the offensive realist state changes its mind-set, then stable peace will be possible. Of course, if the offensive realist state presses on with its aggressive goals, war will still break out.

In summary, while admitting that peace among states tends to be frag-ile, defensive realism does believe that peace has its own dynamics. For defensive realism, war is either the work of evil (i.e., the work of offensive realist states) or the continuation of bad politics (i.e., states’ exacerbating the security dilemma), while peace is the continuation of good politics. Defensive realism thus sees better chances for stable peace, even though peace depends upon many factors and contingencies.

Neoliberalism’s stand on war and peace is rather close to that of defen-sive realism, although neoliberalism has paid little attention to the prob-lem of war. Neoliberalism, however, goes further than defensive realism.


Neoliberalism implicitly rejects the option of force in international politics and believes that economic interdependence and international institutions have made peace among states more robust.

War and Peace: Specific

Offensive RealismOffensive realism strongly advocates offensive behavior. First, an offensive realist state actively seeks conquest and expansion not only because of fear or insecurity but also for the potential gain in relative power (and thus in security too) from conquest and expansion. This is fundamentally due to the fact that offensive realism believes that ultimately security is only assured by universal domination or hegemony (Fischer 1996, 274–276; Mearsheimer 2001; Zakaria 1998).

Second, an offensive realist state not only reacts to an arms race but also actively initiates arms buildup, regardless of whether the buildup induces an arms race or not, simply because the offensive purpose of the offensive realist state demands it to do so. For an offensive realist state, not to arm is always suboptimal (Glaser 2004, 58).15 Moreover, because an offensive realist state aims for conquest and expansion, it needs to adopt an offensive military doctrine and invest more in offensive weapons than in defensive weapons (relatively speaking, of course), when compared to a defensive realist state.16

Third, for an offensive realist state, defensive alliance against an immi-nent threat represents only a temporary solution to meet its security needs. More importantly, for an offensive realist state, alliance does not have to be always for defensive purpose. If possible, two or more states can collude to conquer and then divide the spoils.17 “Jackals,” which are less powerful offensive realist states, will bandwagon for profit, not just for protection (Schweller 1994).18

Fourth, an offensive realist state expands war aims (Labs 1997). If you have your opponents on the rope, you inflict as much damage on them and extract as many concessions from them as you can. This is simply observing the golden rule of maximizing relative power. For offensive real-ist states, “magnanimity in victory” is meaningless. Because other states are inherently aggressive, forsaking the chance to expand war aims will not change their intentions but merely leave more power to them, thus increas-ing the risk that they will grow strong faster and come back for revenge sooner in the future (Machiavelli 1997 [1532], 5.18).19

Fifth, an offensive realist state constantly thinks about preventive war and launches preventive wars (Copeland 2000; Machiavelli 1997 [1532],


3.12; Snyder 1991, 12). In fact, one can argue that in an offensive realism world all wars are essentially preventive wars. The logic is straightforward. States are inherently aggressive toward each other. Thus, even if a state is not a threat now, it will become one (when it is stronger). If so, then it is better and necessary to take offensive initiatives rather than leave your own fate in the hands of others (Hobbes 1968 [1651], 184–185; Machiavelli 1997 [1532], 3.11). In an offensive realism world, preventive war is thus a must, not an option (Tang 2008a, 461–462). An offensive realist state has to constantly seek and actively grab opportunities to weaken other states and, if possible, to eliminate the possibility that other states will ever threaten it either by becoming a universal empire or a regional hegemon. For an offensive realist state, passive posture of deterrence and defense are not enough (Fischer 1996, 275).

Sixth, an offensive realist state also actively seeks and employs other less violent means to weaken other states. These measures can range from “competitive de-colonization” (Hager and Lake 2000), to “weakening and shrinking an opposing alliance” (Waltz 1979, 118),20 to engineering political instability (e.g., supporting insurgency and secessionism in other states), to sabotaging other states’ economic growth (Mearsheimer 2001), and to encouraging and seducing other states into strategic blunders such as overextension and bleeding (Mearsheimer 2001, 153–156).21

Finally, an offensive realist state rejects behaviors such as moderation and self-restraint, because they cannot serve the purpose of maximizing relative power.

Defensive RealismWhile defensive realists have not been as forthright as offensive realists in specifying their preferred state behaviors, the fact that defensive real-ism generally favors defensive means of self-help is not in doubt. Schweller (1996) was right—Defensive realism (with Waltzian neorealism as a not-so-standard version of it) really does have a clear “defensive” (or “status quo” for Schweller) bias.

First, despite their theories of balancing against power or threat being consistent with both defensive realism and offensive realism; Waltz (1979) and Walt (1987) espoused mostly defensive realism policy recommenda-tions. For them, because balancing is prevalent, states that pursue expan-sion and conquest will be counterbalanced by other states, thus they will eventually be defeated. As such, the best way toward security is not through expansion and conquest, but through moderation and self-restraint. In contrast, despite also recognizing the prevalence of balancing behavior, Mearsheimer (2001, 39–45) still recommends that states pursue (regional or global) hegemony, only more wisely.


Second, and related to the first, unlike offensive realism, which empha-sizes potential gains from conquest and the drawback of not grabbing oppor-tunities for (easy) expansions, defensive realism generally emphasizes the perils of imperial overextension (Jervis 1978; Snyder 1991; Kupchan 1994).

Third, by emphasizing that the security dilemma can only be alleviated but not escaped from, defensive realism not only generally rejects offensive behaviors but also generally recommends moderation, self-restraint, and, ultimately, cooperation (Glaser 1994–1995; Jervis 1978).

Fourth, defensive realism considers an arms race generally subopti-mal when not facing a clear and present threat (Glaser 1992; 1994–1995; 2004). Hence, while a defensive realist state may enter an arms race, it con-siders an arms race to be a rational choice only when facing a state that has shown some signs of aggressive intentions. Moreover, because a defensive realist state does not aim for conquest and expansion, it is more likely to adopt a defensive military doctrine and invest more in defensive weapons than in offensive weapons (relatively speaking, of course) when compared to an offensive realist state.22

Fifth, for defensive realism, alliance is almost exclusively for defensive purpose.

Sixth, although preventive wars are also a potential option when facing a (potential) threat, defensive realism does not consider such an option to be a wise choice, unless under extreme circumstances (chapter 3). In other words, the strategic goal of a defensive realist state is mostly for deterrence and defense, rather than expansion through offense.

Seventh, defensive realism emphasizes the drawback of “expanding war aims” (e.g., by inflicting as much pain upon one’s opponent as possible) and instead advocates “magnanimity in victory” so that the two former adversaries can reconcile and build a stable (and hopefully lasting) post-conflict peace (Trachtenberg 2003, 162–164).

Finally, in order to reduce the danger of war, defensive realism can even take unilateral actions to reduce international tensions (see chapter 5 for a more detailed discussion). A classical approach of this type is the approach of “gradual reciprocation in tension-reduction” (GRIT) advocated by Osgood (1962) and Etzoini (1962).

Summary

Overall, while offensive realism and defensive realism do recommend cer-tain common measures when a state faces a clear threat, offensive realism explicitly and implicitly rejects many of the measures that are usually favored by defensive realism (e.g., moderation, cooperation). Defensive realism also explicitly and implicitly rejects many of the measures that are usually favored


by offensive realism (e.g., actively seeking conquest and expansion). Each deems some of the measures favored by the other as counterproductive, if not dangerous, for states’ security. Their differences are summarized in table 4.1.

Defensive Realism on Conflict and Cooperation

General

To discuss conflict and cooperation, let us recall two major points from chapter 2. First, conflict of interest does not amount to actual conflict. Conflict of interest merely means that there is divergence between two states’ interest, whereas actual conflict means that the two states are actively

Table 4.1 State behaviors in war and peace: Offensive realism versus defensive realism

Behaviors that are consistent with both realisms (when a state faces a clear threat)

Behaviors that are only consistent with offensive realism (thus rejected by defensive realisms)

Behaviors that are only consistent with defensive realism (thus rejected by offensive realisms)

Internal balancing through economic growthInternal balancing through arms buildup when facing a clear and present danger

Arms buildup no matter what, with a strong preference for offensive strategies, doctrines, and weaponries

Arms buildup only when facing an imminent threat, with a less robust preference for offensive strategies and doctrines

External balancing by building alliance when facing a clear and present danger

Offensive alliance for conquest and expansion

Alliance are for defensive purpose only

Bandwagoning for security when allies are not available

Bandwagoning for profit

Expanding war aims “Magnanimity in victory”Less violent means of weakening other states (e.g., sabotaging, engineering political instability)

Attempt to change the mentality of the opponent by signaling benign intentions through GRIT

States have to think of preventive wars constantly.

Preventive war under only the most extreme circumstances

Actively seeking expansion and conquest

Moderation, self-restraint, cooperation


engaged in an armed struggle, or at least, are considering the option of force. Second, conflict of interest has both a subjective dimension and an objective dimension, and it can be both reconcilable and irreconcilable. When there is genuinely irreconcilable conflict of interest between two states, actual conflict between them is highly likely. When the conflict of interest between two states is genuine but only illusorily irreconcilable (thus objectively reconcilable), however, actual conflict between them is not inevitable, because cooperation is a viable alternative to actual conflict (Glaser 1994–1995; Jervis 1999, 49–50). Hence, while some of the con-flicts of interest may lead to actual conflicts, others may not. The outcome depends on the nature of the conflict of interest at hand, states’ strategic choices, and the interactions of their choices.

Offensive RealismIn an offensive realism world, every state intentionally threatens other states’ security. As a result, conflict of interest among states is not only genuine but also genuinely irreconcilable. Offensive realism believes not only that the nature of international politics has always been (almost) completely con-flictual, but also that conflict is necessary in international politics (“either I kill you or you will kill me”). There is very little or no common interest among states in an offensive realism world, other than temporary mutual need when facing a common enemy. As such, for offensive realism, there is no possibility of genuine cooperation other than temporary alliances (Tang 2008a, see also Mearsheimer 2001, Chapter 1). Consequently, a state must dedicate all of its resources to prepare for the inevitable conflict.23 Offensive realism further believes that it is better to take initiatives whenever possi-ble. An offensive realist state thus does not shrink from initiating a conflict when it deems the conflict to be (potentially) profitable: preventive war is a must rather than an option (Tang 2008a).

Defensive RealismIn contrast, while defensive realism also believes that the nature of interna-tional politics is fundamentally conflictual and conflict of interest among state is often genuine, it does not believe that all genuine conflicts of inter-est are genuinely irreconcilable. Defensive realism believes that only some genuine conflicts of interest are irreconcilable and that some conflicts of interest, despite being genuine, are reconcilable. Consequently, defensive realism does not believe that states must end up in actual conflicts when-ever they have a genuine conflict of interest at hand. In other words, defen-sive realism believes that in a defensive realism world where most states do not threaten each other’s security intentionally, at least some conflicts are avoidable and unnecessary (Jervis 1999, 49; see also chapter 1).


Most importantly, in a defensive realism world, the need to avoid unnec-essary (and costly) conflicts represents the predominant form of common interest among defensive realist states. Consequently, cooperation becomes a viable and quite possibly a necessary means of self-help. States now need to cooperate, if only for the minimal goal of avoiding costly conflict and preventing mutual harm. When facing a like-minded defensive realist state, a defensive realist state not only does not deem cooperation as inherently risky, but it actually believes cooperation to be potentially profitable or even optimal. A defensive realist state deems cooperation to be inherently risky only when it is sure that it is facing a clearly offensive realist state.

Moreover, defensive realism believes not only that defensive realist states should cooperate under many circumstances, but also that defen-sive realist states can indeed overcome the obstacles posed by anarchy to achieve cooperation under many circumstances (Glaser 1994–1995; Jervis 1999; Kydd 1997a; 2000a). More specifically, defensive realism believes that two defensive realist states can achieve cooperation and avoid conflict if (and perhaps, ultimately, only if) the two states can take measures to signal each other’s true benign intentions, recognize each other as fellow like-minded defensive realist states, and pursue cooperation reciprocally (see chapter 5).

Defensive realism thus does not submit to offensive realism’s position that cooperation is made difficult simply because of (defensive realist) states’ concern for relative gains, incentives to cheat, and the high cost of being cheated. Instead, defensive realism contends that states’ concern for relative gains is not absolute, but conditional. Under many circumstances, states can overcome the concern for relative gains and reach cooperation (Glaser 1994–1995). Likewise, defensive realism contends that the incen-tives to cheat can be contained by the long shadow of the future (Axelord 1984). Indeed, defensive realism actually argues that under certain circum-stances, states may have more incentives to cooperate exactly because the cost of being cheated and mutual defection is high (Jervis 1978, 179–183; see also Kydd 2005, 19, 98).

Finally, and no less critically, defensive realism also recognizes that pursuing cooperation has another invaluable role to play in international politics. When facing a state with unknown intentions, extending an invi-tation to the state to cooperate is a powerful means to gauge its intentions (see chapter 5).

Defensive Realism’s Approaches toward Conflict and Cooperation

Because offensive realism sees little common interest among states and (almost) no possibility for sustainable cooperation, it essentially rejects


cooperation as a viable external means of self-help. The operational code for offensive realism on the issue of conflict and cooperation is thus very simple and straightforward: Cooperation is difficult (if not impossible) and even dangerous, and conflict of interest will have to be resolved by force or the threat of force.

In contrast, defensive realism’s operational code on conflict of interest and cooperation is more complex. In principle, to cope with any conflict of interest with another state, a defensive realist state has to perform two essential tasks before deciding on specific strategies. The first task is to understand the nature of the conflict of interest along the two dimensions identified in chapter 2, genuineness of the conflict of interest and (in)compatibility of the conflict of interest (see figure 2.2 for details).24 The second task is to understand the causes of the conflict of interest. These two tasks are inherently linked and must be performed together.25 Only after having performed these two tasks and reached some tentative conclusions on them can the defensive realist state choose a fitting strategy for coping with the conflict of interest at hand. Fundamentally, a defensive realist state takes the question of conflict and cooperation as an essentially empir-ical question.

Understanding the Nature of the Conflict of Interest at HandFor the first task, the defensive realist state needs to ask two questions. First, is the conflict of interest at hand objectively genuine or objectively illusory? Second, is the conflict of interest at hand irreconcilable or recon-cilable, both objectively and subjectively? Here, I shall illustrate the com-plexity of this process with a brief discussion on the two quadrants given in figure 2.2B.

The upper left quadrant of figure 2.2B depicts a scenario in which the conflict of interest at hand is real but, at the same time, both objectively and subjectively reconcilable. Under this scenario, actual conflict is highly unlikely, most critically because both sides recognize the situation fairly correctly and neither harbors malign intentions. The security dilemma operates under this scenario but may well be dormant. The power compe-tition that is necessary for activating the dormant security dilemma does not have to take place. Nonetheless, actual conflict is still possible because the security dilemma can be easily activated by one side’s actions or mis-perceptions and the security dilemma may then propel the two sides into actual conflict.

If a defensive realist state determines that it is in a situation as depicted by the upper left quadrant of figure 2.2B, then it is also close to believing that the other state is a like-minded defensive realist state. More impor-tantly, the other state also reads the situation quite accurately. In such a


situation, the task for these two states is relatively straightforward (although by no means easy). Specifically, the two states should work together to prevent the dormant security dilemma between them from being exacer-bated by avoiding provocative behaviors (e.g., nationalistic rhetoric, arms buildup along the border between them). Moreover, the two states should pursue extensive cooperation without too much hesitation. When neither side harbors malign intentions toward each other and the conflict of inter-est between them is genuinely reconcilable, there must be a future for the two on the Pareto frontier.

The upper right quadrant of figure 2.2B depicts a situation in which the conflict of interest at hand is real and, at the same time, objectively reconcilable but subjectively irreconcilable. This situation can be divided into two sub-scenarios.

In the first sub-scenario, one or both states mistakenly believe that their interests are irreconcilable (i.e., the irreconcilability or incompatibility is due to misperception and thus illusory), but neither harbors malign intentions. If so, a genuine security dilemma operates. Here, the subjective (i.e., illusory) irreconcilability may be present at the very beginning or be the product of one or both sides’ mistaken belief that is induced by the spiral dynamics of a security dilemma. In this situation, actual conflict is not unlikely but highly avoidable through cooperation, and much depends on whether the two sides can recognize that their interests are not genuinely incompatible and take steps to signal and learn each other’s (benign) intentions.

In the second sub-scenario, the subjective irreconcilability arises because one or both sides intentionally define their interest in an expansive man-ner, thus intentionally creating the subjective irreconcilability. If this is the case, one side’s intentions may be already malignant and the security dilemma will not apply. Actual conflict is more likely in this sub-scenario than in the first sub-scenario.

At this juncture, the defensive realist state needs to ask the second ques-tion: what causes the conflict of interest at hand?

Understanding the Causes of the Conflict of Interest at HandFor the second question, the defensive realist state asks: Given the nature of the conflict of interest, what causes the conflict of interest and gives it its present nature? Is the conflict of interest caused by one’s own expan-sive definition of interest and/or by another state’s expansive definition of interest, and whether such an expansive definition of interest is due to inherently aggressive intentions, opportunistic greed, or unintended miscalculation?26

If the conflict of interest is caused by one’s own expansive definition of interest, then a defensive realist state needs to moderate its definition


of interest accordingly. Unfortunately, even for a genuine defensive real-ist state, this kind of soul-searching is by no means easy. Most states tend to blame others for the existence of a conflict of interest and its apparent incompatibility, due to ethnocentrism.

If the conflict of interest at hand is caused by another state’s expansive definition of interest, then the defensive realist state needs to convince the other state that it needs to change the definition of its interest in order to avoid actual conflict.

More importantly, the defensive realist state must signal its benign intentions and, at the same, try to determine whether the other state is inherently aggressive, by trying to gauge the other side’s intentions. Does the other state define its interest expansively out of misperception, oppor-tunistic greed, or inherent greed? If the other state defines its interest expansively out of misperception or opportunistic greed, then it is more likely to be a defensive realist state. In contrast, if the other state defines its interest expansively out of inherent greed, then it is more likely to be an offensive realist state.

If the other side is a defensive realist state, then one (a defensive realist state) should refrain from actions that may exacerbate the security dilemma. Instead, one should try to convince the other side (a fellow defensive realist state) that it is mistaken in defining its interest so expansively and it has to change its definition of interest or preferences for outcomes in order to avoid conflict. Moreover, one must also try to assure the other side that the conflict of interest between them can be resolved through cooperation (because the conflict of interest is objectively reconcilable) and that one sincerely intends to resolve the conflict of interest through dialogue and compromise.

If the other side is more like an offensive realist state, then while one (a defensive realist state) should still refrain from actions that may exacerbate the spiral between itself and the other side, it must also take some pre-cautionary measures. One must maintain a robust deterrence and defense posture. At the same time, however, one must also signal its own benign intentions to the other side and, more importantly, try to change the lat-ter’s expansive definition of interest. Apparently, if the other side keeps pressing ahead with its expansive goals (i.e., it proves to be an incorrigible expansionist power like Hitler’s Germany), then the conflict of interest has to be settled by actual conflict.

Summary: Specific Strategies for Pursuing CooperationThe first wave of theories of cooperation, despite being overly structur-alistic and neglecting the actual process of seeking cooperation almost entirely, laid part of the foundation for understanding cooperation in


international politics (Baldwin 1993; Jervis 1982; Oye 1986; Milner 1992).27 After that, however, it seems that studies on cooperation within the realism paradigm have largely stalled. Whereas neoclassical realism has produced many studies of war that are both theoretically and empirically rich (reviewed in Rose 1998; Schweller 2003; see also contributions in Lobell, Ripsman, and Taliaferro 2009), there has been a glaring dearth of empirical studies of cooperation that are conducted explicitly through the prism of defensive realism, other than several important contributions at the abstract level (e.g., Glaser 1994–1995; Jervis 1999; Kydd 1997b) and a few notable exceptions of empirical studies (Kydd 2005; Larson 1997; Ripsman 2005).28

Hence, within the realism literature, there is really very little stuff for this author to go on in order to extract some general principles for states to pursue cooperation. For instance, although there have been many confidence-building measures (CBMs) among states (especially after the cold war) and CBMs obviously fall within the domain of defensive real-ism (because CBMs are about military or security cooperation, a domain that defensive realism has traditionally claimed as its exclusive domain), only practitioners and policy analysts have written extensively on CBMs. In contrast, theorists of defensive realism have pretty much ignored this important aspect of security cooperation altogether!29

As such, I shall end this section with an explicit call for more empiri-cal studies on cooperation within the defensive realism paradigm, if defen-sive realism insists that cooperation is possible within defensive realism. Otherwise, defensive realism may simply want to cede the territory of cooperation to neoliberalism or constructivism. The debate on cooperation among the three grand theories of international politics has stayed at the abstract and structural level for way too long, and it is time for a change.

The General Operational Code of Defensive Realism

After the preceding discussions, we can now outline the general opera-tional code of defensive realism.

Differentiating an Offensive Realist State from a Defensive Realist State

For a defensive realist state, the first thing to do when formulating strat-egies is to differentiate states into two categories: offensive realist states versus defensive realist (and neoliberalist) states.30 This process of differ-entiating states or determining the nature of other states is of paramount importance for a defensive realist state (Jervis 1999, 51–52). Indeed, unless


this question is answered (however tentatively), a defensive realist state will be at a loss for making strategies.

On the one hand, if a defensive realist state considers another defensive realist state as an offensive realist state, it will play too much of a deterrence game but little of an assurance/cooperation game (Jervis 1976, Chapter 3; Kydd 1997b; 2000), and thus may exacerbate the security dilemma between itself and the other like-minded defensive realist state, eventually ending up in an unnecessary arms race and conflicts. On the other hand, if a defensive realist state takes an offensive realist state as a defensive real-ist state, the defensive realist state will play too much of an assurance/cooperation game but too little of a deterrence game, and thus may simply provide an offensive realist state with a much-coveted opportunity to strike a decisive or even fatal blow upon the defensive realist state. In sum, if a defensive realist state makes a mistake in assessing the nature of another state, it will commit strategic blunders both ways: either doing too much or doing too little.

So, what are the criteria that a defensive realist state can employ to differentiate a like-minded defensive realist state from an offensive realist state? Kydd suggested four criteria, and they are: ideology, policy toward domestic minorities, policy toward weaker neighbors, and military policy and arms control (Kydd 1997b, 141–147). I believe that the following two criteria, which cover both words and deeds, are more suitable for differenti-ating an offensive realist state from a defensive realist state, and these two criteria subsume Kydd’s criteria.

The first criterion is whether a state recognizes the prevalence of bal-ancing behavior and the existence of the security dilemma (and spiral) and grasps at least some of the two phenomena’s moderate (i.e., defensive) implications for state behavior. A defensive realist state does recognize this while an offensive realist state does not.31 This criterion is essentially to measure whether a state understands the logic behind the necessity for moderate behavior.

For instance, a defensive realist state understands that states cannot escape from the security dilemma simply by accumulating more and more power, and that states can only try to alleviate the security dilemma through moderation and cooperation. In contrast, an offensive realist state either denies the existence of the security dilemma or draws the lesson that one has to try to escape from it. Similarly, a defensive realist state understands that the prevalence of balancing behavior means that aggres-sive actions usually promote counterbalancing and leads to a decrease of security. In contrast, in light of the prevalence of balancing behavior, an offensive realist state draws the implication that one simply has to pursue aggression more wisely.


The second criterion is whether a state exercises self-restraint and is willing to be constrained by other countries. A defensive realist state exer-cises self-restraint and is willing to be constrained, and an offensive realist states is not willing to do so (see discussion below). Because this pattern of behavior is inherently tied to the understanding of the logic behind the necessity of moderate behavior, this criterion thus measures not only whether a state really behaves with moderation but also whether it takes the logic behind the necessity of moderate behavior to its heart. As folk wisdom has it, one uses deeds to measure words.

The second criterion applies mostly because an offensive realist state has to constantly seek and grab opportunities of weakening others in order to maximize its relative power versus other states, whereas a defensive realist state does not do so because it does not believe that maximizing relative power is always a good means toward security. In order to achieve its expan-sive goals, an offensive realist state cannot afford to exercise self-restraint and to be restrained by others. In contrast, because a defensive realist state does not pursue expansive goals, it can afford to exercise self-restraint and be restrained by others.

So how does a defensive realist state learn the other state’s true inten-tions? The short answer is this: signaling benign intentions by striking a costly cooperative posture and then extending an invitation to coop-erate to the other state, without risking one’s vital security interest (see chapter 5 for details). If the other state reciprocates, then it is more likely to be a like-minded defensive realist state. If the other state responds by taking advantage of the goodwill of the defensive realist state, then the other state is more likely to be an offensive realist state. Moreover, this method of “giving trust and testing the other side’s intention” can also differentiate a defensive realist state with opportunistic greed from a hardboiled offensive realist state. The former will forsake its occasional greedy goals and can be persuaded to change its mind-set with words backed by power, whereas the latter will consistently pursue greedy goals and can only be persuaded to change its mind-set by raw power. This simple measure of extending an invitation for genuine cooperation and then reading the other side’s reaction not only makes intuitive sense, but also has firm foundation in the theory of trust and mistrust. Indeed, because it makes intuitive sense, statesmen actually deploy the measure in the real world.32

A defensive realist state thus tries to read other states’ behaviors con-stantly in order to update its assessment of other states’ true intentions and then adjusts its strategies accordingly (Kydd 2005, 26–27). In con-trast, because an offensive realist state believes that states are inherently aggressive, it does not have to update its assessment of the other states’ true


intention, nor does any updating impact its policies in any meaningful way (Kydd 1997b, 126; 2005, 14).

General Strategies: The Military Front

After a defensive realist state is able to roughly tell whether a state is a like-minded fellow defensive realist (or even neoliberalist) state or an offensive realist state, the defensive realist state can then apply a general strategy in dealing with the other state. Needless to say, the strategy toward a like-minded fellow defensive realist state is quite different from the strategy toward an offensive realist state, on both the military front and the polit-ical front.

When Facing an Offensive realist stateWhen confronting an offensive realist state, a defensive realist state adopts a security strategy that may seem somewhat similar to that of an offensive realist state when looking from the point of view of an arms race. Such a strategy is the defensive realist state’s rational response when facing a clear and present threat (Glaser 2004).33

The military dimension of such a security strategy consists of two major components. First, the defensive realist state strives to maintain a robust deterrence and defense posture against the offensive realist state by vigor-ously arming itself. When facing an offensive realist state, an arms race is rationale and thus inevitable. The (cold war) hawks do have a point: When facing an offensive realist state, arms control will be difficult, if not impos-sible, to achieve (Gray 1993; Kydd 2000b).34

Second, the defensive realist state also strives to form an alliance (not necessarily with like-minded defensive realist states) to defensively balance the threat (Walt 1987).35 The primary objective of such a security strategy is to prevent the offensive realist state from attacking and to prevail in the conflict. Failing to adopt either of the two measures on the part of the defensive realist state will lead to suboptimal internal and external balanc-ing (Glaser 2004; Lobell, 2002–2003; Schweller 2006).

There are, however, two significant differences between the military strategy of a defensive realist state and that of an offensive realist state, even when both of them are facing another offensive realist state.36

First, the military strategy of a defensive realist state is deterrence and defense, but that of an offensive realist state is expansion. A defensive real-ist state usually chooses to live with an offensive realist state (and eventu-ally change its mentality), rather than seek opportunities to wipe it out. In other words, a defensive realist state does not consider preventive war as a valid option, unless under extremely rare circumstances. More precisely,


facing an offensive realist state, a defensive realist state will generally adopt a strategy between engagement and passive containment in the ladder of strategies (figure 4.1), but will rarely adopt strategies that are more con-frontational than active containment.

In contrast, an offensive realist state actively considers and may actually launch preemptive war, preventive war, and opportunistic war of expan-sion. An offensive realist state seeks and grabs windows of opportunity to gain upon another state, regardless of whether the other state is an immi-nent threat or not. Conciliatory measures such as engagement and passive measures such as passive containment (i.e., pure deterrence/defense) are not enough for offensive realist states, simply because they do not serve the offensive realist state’s purpose of expansion and conquest.

Second, even when facing an offensive realist state, the military strategy of a defensive realist state is not, as many tend to believe, a simple turn to a simplistic and hard-line policy of arms race and deterrence. Instead, a defensive realist state crafts a calibrated combination of both deterrence/defense and assurance to handle the situation. Insights from the spiral model point to a more nuanced policy of simultaneously reducing the inse-curity of the offensive realist state while making it clear that aggression does not pay and moderation is the way to go (Glaser 1992; 1994–1995; Kydd 1997a; Snyder 1985). Such a policy requires the defensive realist state to reassure the offensive realist state by signaling benign intentions while at the same time maintaining a robust deterrence and defense posture (see also chapter 5).

In this strategy of engagement, the military posture of a defensive real-ist state serves as a powerful reassuring gesture to the greedy but still inse-cure offensive realist state that the defensive realist state is not malignant.37 When a defensive realist state understands the political consequences of military strategy (Glaser 1992), it can forsake unnecessary offensive mili-tary capabilities, pursue unilateral defense, and exercise unilateral restraint in order to reassure the offensive realist state (see chapter 5). Consequently, even when facing an offensive realist state, the military strategy of a defen-sive realist state itself is a calibrated combination of deterrence and defense on one side and assurance on the other (Snyder 1985, 177–179; Christensen 2001; 2002, 8–10).38

In contrast, even when not facing another offensive realism threat, an offensive realist state accumulates as much offensive capabilities as possible and does not exercise unilateral restraint or a purely defensive posture, sim-ply because its offensive goals require it to do so (Glaser 1992; 1994; Kydd 1997b). For an offensive realist state whose goal is security through dom-ination, armaments dictated by the necessity of deterrence and defense simply are not adequate for its aggression and expansion.


In other words, even when engaging in an arms race, a defensive realist state does it differently from an offensive realist state. Fundamentally, a defensive realist state takes arms race as a means toward an eventual peace-ful resolution of a conflict of interest, without actually resorting to actual conflict. An offensive realist state essentially resorts to arms race purely as a means toward compelling its opponent into submission or for the neces-sary preparation for eventual aggression.39

Overall, even when facing a (potential) offensive realist state, defen-sive realism believes that the combination of too much (thus unnecessary) offensive military power and too little signaling of benign intentions is more likely to be counterproductive than the opposite combination. Of course, a defensive realist state does not retreat when an offensive realist state keeps pressing on with its expansionist goals. A defensive realist state is prepared to stand firm and fight if necessary.

When Facing a Defensive Realist StateWhen facing another like-minded defensive realist state, the military strat-egy of a defensive realist state has to serve two fundamental goals.

The first is to mitigate the security dilemma, to reduce misperceptions, and to avoid any unnecessary arms race and conflicts between them. This objective is usually achieved through two major means: by exercising self-restraint, and by being willing to be constrained by other countries (Glaser 1992; 1994–1995; Taliaferro 2000–2001; Tang 2004; Wolfers 1952).40 Between two defensive realist states, exercising self-restraint and being willing to be constrained is a powerful tool for containing the security dilemma between them.

The second goal is to further cement the relationship between the two states. This is usually achieved by moving beyond cooperative measures that merely alleviate the security dilemma (e.g., stopping the arms race and initiating arms control) and moving into a higher form of cooper-ation. Most prominently, confidence-building measures (CBMs) are a higher form of security cooperation that two defensive realist states can pursue, short of alliance. Between two defensive realist states, CBMs such as troop reduction along their common border and having an early warn-ing system of troop movement not only allow the two states to reduce each other’s apprehension about the other side’s military maneuvering,41 but also allow the two states to shift military power from their border region to other more needed theaters. Meanwhile, CBMs such as joint exercises and intelligence sharing allow the two states to cooperate with each other more effectively if they were attacked by common enemies in the future. CBMs thus allow the two states to gain security from the deeper and more extensive cooperation between them, both directly and indirectly.42


Fundamentally, what truly differentiates a relationship between two defensive realist states from a relationship between a defensive realist state and an offensive realist state is that, in the latter situation, the defensive realist state should be contended with stopping an arms race and enforcing arms control, whereas in the former, the defensive realist state should move beyond stopping an arms race and enforcing arms control into deeper and more extensive security cooperation.

General Strategies: The Political Front

When Facing an Offensive Realist StateAt the political level, the strategy of a defensive realist state when facing an offensive realist state differs even more significantly from that of an offen-sive realist state when facing a fellow offensive realist state.

The ultimate political goal of the strategy of the defensive realist state is to change the preference for goals (e.g., desiring somebody else’s terri-tory) and strategies of the offensive realist adversary through influencing the domestic debate and inducing changes of political power within the adversary’s society through peaceful means (Evangelista 1993; Glaser 1992, 517–527; Kydd 2000a; Stein 1991). Understanding that there are doves (i.e., defensive realists) and hawks (i.e., offensive realists) within the adver-sary state, the defensive realist state seeks to tilt the political support toward the doves through sending costly reassuring signals that are designed to convey benign intentions and a willingness to cooperate, while maintaining a robust deterrence and defense posture. The ultimate goal is to turn the initially offensive realist state into a defensive realist state.43 After the adver-sary changes from an offensive realist state to a defensive realist state, the defensive realist state can then seek extensive cooperation with the former offensive realist adversary, just as with another defensive realist state.

In this sense, the political strategy of a defensive realist state toward an offensive realist state consists of two distinctive phases. In the first phase, the task for the defensive realist state is to change the offensive realist state into a defensive realist state, so that the situation between the two states moves from a security threat (for the defensive realist state) to a real secu-rity dilemma. In the second phase, the defensive realist state then seeks (extensive) cooperation with the former offensive realist adversary, which is now a like-minded fellow defensive realist state.

Of course, a defensive realist state should have realistic expectations when it comes to changing the intentions of an offensive realist state. The process may take a fairly long time. Worse, the offensive realist state may be an incorrigible aggressor (e.g., a Hitler) and decide to press on with its


aggressive goals. As such, a defensive realist state must also remain vigilant even when pursuing a strategy of engagement. It must be ready and willing to stand firm and fight if necessary.

In sharp contrast, an offensive realist state gives up on seeking changes in other states’ behavior through influencing their domestic politics with peaceful means, because it believes that every state is inherently aggres-sive and thus there is no point in seeking changes to a state’s nature. Alternatively, an offensive realist state seeks to change other states’ behavior by imposing its will upon other states (either directly or indirectly through a puppy regime). Mearsheimer and other traditional offensive realists (i.e., paleo-cons) will recommend the first option, while liberal offensive realists (i.e., liberal hawks or the “neo-cons”), the second (see Appendix II).

When Facing a Defensive Realist StateWhen facing a fellow defensive realist state, the political strategy of a defensive realist state is to cement the relationship between them, and thus make sure that the other state does not change into an offensive realist state. This objective can be achieved through two means.

The first is to consistently behave with moderation toward each other. This is to solidify the support for the voice for good relationship between the two states and simultaneously marginalize the voice for bad relationship within each other’s domestic politics. In other words, this means is to make sure that the hawks in the other country do not gain political power.

The second is to develop and sustain extensive cooperation between the two states. The extensive cooperation will not only bring tangible benefits to the two states, but also serve as a sign of continuous goodwill toward each other. The extensive cooperation of the two states serves as a sort of binding strategy, and may even eventually bring the two states into some sort of a security regime (Jervis 1982), without necessarily bringing the two states into a formal alliance.44

Of course, a defensive realist state should continue to remain vigilant toward a fellow defensive realist state, for its leaders’ (definition of) inter-ests and intentions can change any time. Nonetheless, the extensive coop-eration between the two states and the goodwill accumulated from such cooperation should provide the two states with ample reasons to not over-react toward each other’s moves.

Formulating Specific Policies

When it comes to assessing its security environment and formulating spe-cific security policies, a defensive realist state calculates far more factors


than an offensive realist state does (Buzan 1991, Chapter 2; Glaser 1992, 525–533; 1994–1995; Jervis 1978; 1988; Kydd 1997b; Taliaferro 2000–2001; Tang 2004; Walt 1987). As a result, formulating policies for a defen-sive realist state is a far more complicated process than for an offensive realist state, and it is not sensible to offer a generalization here. Nonetheless, four general points can be made.

First, despite some overlapping, offensive realism generally prefers the upper part of the ladder (i.e., more confrontational strategies), whereas defensive realism generally prefers the lower part of the ladder (i.e., more conciliatory strategies) depicted in figure 4.1. In principle, an offensive realist state does not adopt any conciliatory options beyond active con-tainment. In contrast, a defensive realist state generally does not adopt any confrontational options beyond active containment. However, even a defensive realist state may adopt more hardened strategies such as active containment without giving much thought to engagement when the defen-sive realist state mistakenly believes that it is facing an offensive realist state, as predicted by the security dilemma/spiral model.

Second, even when not facing an imminent threat, an offensive real-ist state still chooses more confrontational strategies, simply because its expansive goals demand it to do so. In contrast, even when facing an immi-nent threat, a defensive realist state will generally refrain from choosing preventive war, the most confrontational strategy.

Third, when facing a possible threat, both offensive realism and defen-sive realism will reject appeasement and do-nothing. For realism, appease-ment and do-nothing are irrational behaviors, unless under extreme circumstances (e.g., when a state faces an aggressor but the state alone is too weak to resist the aggressor and no allies are available). As such, these behaviors will have to be explained by domestic pathologies or poor states-manship, as neoclassical realism has maintained (Rose 1998).

Finally, because even a genuine defensive realist state may occasionally pursue expansive goals out of fear or opportunistic greed (especially when facing another state that is engaging in suspected expansive behaviors), a defensive realist state usually cannot determine with great confidence within a short timeframe whether the other state is a defensive realist state that is engaged in opportunistic expansion or a hardened offensive realist state. Only time will tell, and one has to pay a price to get to know the other side’s intentions (information is costly). Because uncertainty is inher-ent under anarchy, most defensive realist states will do some “hedging” in the real world. However, the more certain a defensive realist state becomes about the nature of another state, the surer its strategy becomes and the less hedging it resorts to (see chapter 5).


Making Strategy: Offensive Realism versus Defensive Realism

For offensive realism, states are inherently aggressive, either by nature or because of anarchy. Offensive realist states thus assume the worst about other states’ intentions and infer other states’ intention exclusively from their capabilities: a state will be aggressive when it possesses the capabil-ity to be aggressive. In contrast, defensive realism explicitly warns against assuming the worst about others’ intentions. Defensive realism advocates

Table 4.2 The general operational codes of the three approaches at a glance

Aspects of Operational Codes

Offensive Realism Defensive Realism Neoliberalism

Are there defensive realist states out there?

No. Yes. Yes.

Should a state try to assess the other state’s intention?

No, because all states are offensive realist states.

Yes, there are both defensive realist states and offensive realist states out there.

Yes, there are both defensive realist states and offensive realist states out there.

General Strategies To constantly seek and take advantage of opportunities to weaken all other states, temporarily pausing only when an alliance is necessary.

When facing like-minded defensive realist states: self-restraint and willingness to be restrained. Seeking cooperation.

When facing like-minded defensive realist states: self-restraint and willingness to be restrained. Seeking cooperation.

When facing offensive realist states: deterrence and defense with arms and alliances; attempt to change the nature of the adversary if possible.

When facing offensive realist states: deterrence and defense with arms and alliances; attempt to change the nature of the adversary if possible.

Moving toward codified cooperation (i.e., institutions)?

No. Yes, institutions are helpful, but they are not necessary.

Yes, institutions are both necessary and helpful.

Behave according to the norms and ideas dictated by institutions?

No, only according to interests.

No, only according to interests.

Yes. Ideas, Norms dictated by institutions are interest themselves.


states to differentiate other states into offensive realism ones and defen-sive realism ones while taking precautionary measures to hedge against the possibility that another state is aggressive. Moreover, defensive realism urges states to constantly update their reading of others’ intentions and act accordingly (see chapter 5; see also Tang 2008a; Glaser 1994–1995). This difference between defensive realism and offensive realism on how to cope with the uncertainty about others’ intentions gives four broader contrast-ing characteristics to the two realisms.

First, defensive realism considers far more factors than offensive realism does when making strategies. For defensive realism, war and peace and conflict and cooperation depend on far more variables than offensive real-ism conceives. As such, defensive realism is less parsimonious than offen-sive realism. Second, because defensive realism considers far more factors than offensive realism when making strategies, defensive realism is more sophisticated than offensive realism.45

Third, defensive realism is a more flexible theory than offensive real-ism. Because a defensive realist state constantly updates its perceptions about others’ intention and acts accordingly, it is less likely to commit the mistake of staying course when facing increasing evidence that its original perception may be wrong. From the perspective of learning, a defensive realist state has lower cognitive inertia than an offensive realist state. This is especially true when facing a state with unknown intentions. In such a situation, while guarding against the possibility that the other state may be an offensive realist state, a defensive realist state also takes measures to sig-nal its benign intention to the other state and, more importantly, to learn the true intention of the other state along the way. In contrast, an offensive realist state does not do that.

Finally, for offensive realism, because states are inherently aggressive, peace is merely the intermission between wars and it has no dynamics of its own. For offensive realism, politics is all about preparing for and win-ning the inevitable war, and war and politics become essentially the same. As such, offensive realism is essentially a theory of preparing and win-ning war. In contrast, defensive realism rejects the notion that states are inherently aggressive (see table 4.2). While admitting that peace among states tends to be fragile, defensive realism believes that peace has its own dynamics and that robust peace between two defensive realist states can be built upon extensive cooperation. For defensive realism, war is either the evil work of offensive realist states or the continuation of bad politics (i.e., exacerbating the security dilemma or the spiral), while peace is the con-tinuation of good politics. As such, defensive realism is not only a theory of preparing and winning war, but also a theory of peace and cooperation-building.


Chapter Five

Reassur ance: A Defensive Realism Theory of Cooper ation-building

The discussion in chapters 2 and 4 makes it evident that the way to cope with uncertainty over others’ intentions constitutes a critical point of di-vergence between offensive realism on the one side and defensive realism (and other non-offensive realism approaches) on the other (Tang 2008a).

Offensive realism’s solution is fairly simple. Offensive realism believes that states are inherently aggressive either by nature or because of an-archy, and/or that their aggressiveness is only limited by their capabilities. Furthermore, (others’) intentions are inherently difficult to gauge (partly because they can change).1 As such, states (have to) assume the worst about others’ intentions, thus essentially eliminating the certainty over others’ intention from their strategic calculus (Labs 1997, 11–12; Mearsheimer 2001, 32–36, 45).2 When states assume the worst about others’ intentions, cooperation other than temporary alliance when facing a common threat becomes extremely difficult, if not entirely impossible (Tang 2008a).

Defensive realism’s approach toward the problem is far more complex. Despite admitting that others’ intentions are difficult to gauge partly be-cause they can change, defensive realism believes that states are not inher-ently aggressive and that genuine cooperation—although often difficult—is possible. More concretely, two benign states can achieve cooperation by re-ducing their uncertainty over each other’s intentions and establishing some trust between them via reassurance. In other words, defensive realism has signaled reassurance as the principal means for unwinding, or at least con-taining, a spiraling security dilemma. “The importance for developing a good theoretical and empirical understanding of reassurance in interna-tional affairs is therefore manifest” (Kydd 2000a, 325–326).

The intellectual foundation of the defensive realism’s approach can be traced back at least to Arnold Wolfers’ formulation that there is a funda-mental trade-off between “efforts to change the intentions of an oppo-nent” and “efforts to build up strength against him” (Wolfers 1952, 497).


Since then, generations of scholars, including Osgood (1962), Etzioni (1962), Jervis (1976; 1978), Glaser (1992; 1994–1995), Stein (1989; 1991), Larson (1997), Lebow (1997), Kydd (1997b; 2000a; 2005, esp. Chapter 7), and Schultz (2005), have all contributed important insights into this tradition. Despite having formulated some powerful ideas about the problem, however, defensive realism has yet to offer a truly coher-ent statement about how a (defensive realist) state should cope with the uncertainty about others’ intentions through reassurance and achieve cooperation— “there is no ‘theory’ of reassurance” (Stein 1989, 24; see also Kydd 2005, 185; Montgomery 2006). Indeed, as it becomes clear below, existing literature suffers from some crippling errors and remains underdeveloped in some critical areas, and much confusion and inconsis-tency remains. Not surprisingly, offensive realists such as Mearsheimer (2006b, 232–234) can continue to taunt his opponents: “show me how [states can cooperate].”

Critically building upon existing literature, this chapter seeks to ad-vance a more coherent statement about reassurance, the core component of a defensive realism’s approach toward coping with the uncertainty about others’ intentions, thus moving us a step closer to a defensive realism theory of cooperation-building.3 Because cooperation-building via reassurance is essential for forging the initial foundation for more institutionalized co-operation later (Tang 2008a, 462–465), a more coherent understanding of reassurance also contributes to the understanding of building international cooperation in general. I advance three key arguments.

First, reassurance is fundamentally a means of gradually building co-operation and accumulating trust in order to eventually move one’s rela-tionship with others into a more cooperative state.4 Second, reassurance is simultaneously a means to signal one’s own benign intentions and a means to read others’ present intentions by reading others’ reactions to one’s invi-tations to cooperate,5 with “costly signaling” under risk at its core. Third, and related to the first, piecemeal cooperation builds trust, and more trust leads to more extensive cooperation. Trust and cooperation have a feed-back rather than a one-way relationship.

Before I proceed further, three caveats are in order. First, although costly signaling with its inherent risk lies at the heart of reassurance, reassurance cannot possibly be a purely “rational choice” phenomenon because trust and fear inherently have a psychological/noncognitive/emotional dimen-sion and a “rational”/cognitive/strategic dimension (Lewis and Weigert 1985). In order to understand reassurance, we have to integrate the psycho-logical approach and rational choice approach. Second, although a states’ strategic environment profoundly impacts reassurance, I shall mostly limit my discussion to “simplified” conditions to make things tractable.6 Third,

theory of cooperation-building / 131

and perhaps most important, although I am deeply aware that domestic politics is crucial for understanding cooperation-building in the real world, I refrain from engaging it here other than emphasizing the necessity for more work on this topic in the conclusion. We simply know too little about the domestic politics of cooperation-building, comparing it to the growing body of neoclassical realism literature on the domestic politics of conflict (Schweller 2003; Lobell, Ripsman, and Taliaferro 2009).

The rest of the chapter is divided into five major sections. Section 1 defines concepts more rigorously and clarifies the relationship between reassurance and several other related concepts. Section 2 highlights some major errors and areas of underdevelopment within existing litera-ture. Section 3 offers a more comprehensive categorization of reassurance attempts. Section 4 focuses on conditions and factors for initiating, recip-rocating, and succeeding with reassurance. Section 5 briefly restates the core logic of reassurance and provides a brief guide to practice reassurance. A concluding section follows.

Definitions and Clarifications

Uncertainty, Intentions, Trust, and Reassurance

In addition to the external factor of “strategic environment,” four internal factors that interact with each other, namely (military) capabilities, inten-tions, interests, and resolve (to fulfill one’s commitments), are critical for understanding and anticipating states’ behavior, and uncertainty can form over any one of them.7 Among the four internal factors, uncertainty over others’ capabilities and interests is perhaps the least problematic in terms of detection, because capabilities and interests can be more readily observed or fathomed. Uncertainty about others’ resolve generally weighs in only when two states are already involved in serious bargaining. In contrast, uncertainty about others’ intentions poses a problem for a (benign) state from the very beginning when it seeks to forge a sound security strategy (see chapter 4 above). Apparently, uncertainty about others’ intention can arise under many circumstances, such as change in strategic environment (e.g., the end of the cold war), change in relative capabilities, and change of leadership in the two states.

An actor’s intentions are his preference over actions (for our purpose here, benign or malign) for achieving his goals. An actor’s motives are immediate payoffs or goals that drive an actor’s behavior, and thus one’s more immediate interests. Intentions are thus “preferences over strategies” whereas motives are “preferences over [immediate] outcomes” (Powell 1994; see also Jervis 1976, 48–49).8


Trust is a function of one’s estimation that the other individual or state is benign (i.e., it will not intentionally threaten one’s life or interest). The opposite of trust is mistrust or distrust, and mistrust is a function of one’s estimation that the other individual or state is malignant.9 Trust and mistrust thus are two poles of the continuous spectrum that is par-tially underpinned by one’s uncertainty over others’ intentions, and they have an inverse relationship: trust reduces mistrust, and vice versa. The middle point of the spectrum is non-trust (i.e., neither trust nor mistrust, see figure 5.1). Apparently, more trust leads to a stronger sense of security from others whereas more mistrust leads to a stronger sense of insecurity from others, or fear of others, assuming that others possess capabilities to do harm.10

Reassurance, Cooperative Strategies, Trust, and Cooperation

To reassure is to “say or do things to make others stop worrying about something.”11 Translated into the language of international politics, to reassure is to make other states worry less about one’s intentions (and hence one’s capabilities too) because one’s intention is genuinely benign. Reassurance is thus a specific form of signaling: it is to signal one’s benign intentions toward another state. Obviously, as a signaling process, a reas-surance attempt must have a sender and a receiver.

Reassurance is also a means to the end of knowing the other side’s inten-tions. One signals one’s benign intentions by issuing a cooperative gesture and extending an invitation toward cooperation. Afterward one can gauge other states’ intentions from their reactions toward one’s goodwill gesture

Trust Mistrust

Sense of security Fear or sense ofinsecurity

Reassurance andcooperation

Provocation andcompetition

Uncertainty overothers’ intentions

Others’ capabilities todo harm

Figure 5.1 Trust versus mistrust.


(Kydd 2005, 19–20, 184–18l; Schweller 1999, 14–15; see also Weinberger 2002).

For cooperation-building, reassurance serves two interconnected func-tions. The first function is to assure other states of one’s own benign inten-tions by issuing a cooperative gesture and extending an invitation toward cooperation and to simultaneously gauge other states’ intentions from their reactions toward your reassurance attempts. The second function, which builds upon the first function, is to move the bilateral relationship be-tween two states into a more cooperative mode by increasing their mutual trust or reducing their mutual mistrust so that they can behave benignly toward each other, when they both are genuine benign states.12 Together, this means that reassurance is part of a process of building more trust through forging gradual cooperation so that the level of cooperation be-tween two benign states can be gradually increased or the level of conflict can be gradually reduced (see section 2 for details).

Normally, to move a bilateral relationship one step toward a more co-operative relationship, or even to determine the true intentions of another state, requires a series of, rather than a single, reassurance attempts. A series of reassurance attempts is a reassurance program. Obviously, a reassurance program can ultimately succeed even if some of the reassurance attempts within the program fail, although a reassurance program cannot succeed when all its attempts fail.

A reassurance strategy, usually manifested in a reassurance program, is usually not an independent strategy by itself. Because pure reassurance is too risky (i.e., the other side may be malign), states almost invariably deploy reassurance in combination with deterrence/defense. Thus, a reas-surance strategy is usually an integral part of a broader cooperative strategy called engagement, which takes place between two states that are unsure of each other’s intentions. Engagement has both a reassurance element and a deterrence/defense element (Schweller 1999, 14–15; Stein 1991, 432–433, 451). In other words, at the heart of engagement there is an inherent hedg-ing element and a difficult trade-off between reassurance and deterrence/defense.

For two former adversaries, a reassurance program is part of a larger process more generally known as “reconciliation” or “peace-building/mak-ing” (Akerman 1994; Miller 2000). Reconciliation is thus a special kind of reassurance program after a major war or a major conflictual phase of a bilateral relationship. Obviously, reconciliation is far more difficult than regular “cooperation building,” simply because reconciliation occurs after a major conflict.

Cooperation or collaboration is different from coordination or “spon-taneous cooperation” (Stein 1982). Coordination is a situation in which


two parties with a common aversion “spontaneously” reach a tacit under-standing to avoid the undesired outcome, without their intentions being necessarily benign toward each other. The emergence of the norm of avoid-ing direct conflict and nuclear war between the two superpowers during the cold war was a case of coordination at the state level, whereas the “live-and-let-live” attitude between the two opposing armies during the trench war of WWI was a case of coordination at the substate (tactical) level. In both cases, the two opposing sides did not change their mind-set against each other. They still wanted the other side to eventually lose the conflict, and they managed to avoid a disastrous outcome only because such an out-come would be a disaster to both sides. In other words, both sides still play a zero-sum game rather than a total-sum game in coordination, and there is no trust (in each other’s benign intentions) involved. Both sides are con-fident that the other side will not defect, simply because they are confident that defection will mean a disaster for both.

In contrast, cooperation requires a conscious effort to bridge the con-flict of interest and strive for achieving common interest. Cooperation is thus a total-sum game (Keohane 1984, 53–54; Milner 1992, 467–470).13 More critically, cooperation requires some trust ex ante, but it also builds more trust into a relationship (see section 2 for details).

Cooperation is not to be confused with cooperative strategy or behavior. First, cooperative behaviors or strategies do not necessarily lead to coopera-tion. Whereas cooperative behaviors or strategies can be a one-sided effort, cooperation is an outcome that can only be engineered by both sides. Thus, reassurance is a form of cooperative behavior or a part of a (cooper-ative) strategy, but it is not cooperation itself. Second, strategies other than cooperation can also be considered as cooperative strategies. Cooperative strategy is thus broader than cooperation.14 For instance, whereas both appeasement and engagement can be considered as cooperative strategies toward another state, only engagement can lead to genuine cooperation. Chamberlain certainly wanted to achieve some kind of accommodation with Hitler, but he failed miserably because Hitler wanted nothing to do with cooperation. In contrast, engagement between China and Southeast Asian states in the past two decades or so has resulted in some genuine cooperation between the two sides (Collins 2000, Chapter 5; Shambaugh 2004–2005; Glosny 2006).

Existing literature also tends to employ cooperation to describe a mode of state relationship. To avoid confusion, I use “cooperative relation-ship” to describe a relationship that has accumulated a significant level of cooperation.

Finally, it is necessary to differentiate a genuine reassurance attempt from a false reassurance attempt. Unlike a genuine reassurance attempt


that signals genuine benign intentions, a false reassurance attempt is a ploy designed to lull other states into a false sense of security even though one’s intentions are malign. Alternatively, a false reassurance attempt is a pub-licity stunt to show one’s domestic audience and/or other states that one is flexible and benign whereas one’s adversary is intransigent and malignant. Typically, for a false reassurance attempt, one makes a small gesture and demands a (far) costlier gesture from one’s adversary in return, knowing fully well that this demand will be rejected. In both cases, one has no real intention to move the bilateral relationship to a more cooperative mode. Hitler’s 1939 nonaggression pact with Stalin was a classic example of the first kind of ploy. George Kennan’s extending economic aid to Eastern European states and the Soviet Union in order to strain the relationships between the Soviet Union and its satellites was a classic example of the second kind (Gaddis 2005, 37, 56).15

Cost, Risk, and Credibility

Existing literature has not been explicit or consistent with three key terms for understanding reassurance: cost, risk, and credibility of a reassurance attempt or signal. The cost of a reassurance signal is relatively straightfor-ward to understand. The cost can be either direct or indirect, and it can involve either (direct) material cost incurred by the signal or symbolic cost such as losing some of one’s domestic standing when the signal is assailed by one’s domestic opponents as “appeasing.”

The difficulties lie with understanding a signal’s risk and credibility. As stated above, a reassurance signal is essentially an invitation to reciprocate one’s own cooperative gesture. Because a reassurance signal inevitably puts some of one’s welfare in the other side’s hands and there is always the pos-sibility that the other side may not reciprocate, a reassurance signal always entails some risk of losing some of one’s welfare (Deutsch 1958, 265–267; Gambetta 1988; Giffin 1967, 104–105; Hardin 2002, 11–13, 113–129; Larson 1997, Chapter 1).16 One will not initiate a reassurance signal if he/she is unwilling to take this risk, and the signal—even if initiated—will not carry any credibility if there is no real risk in issuing the signal (Cook et al. 2005, esp. 125—26, 140; Yamagishi et al. 2005, 276—79; Kydd 2005, 41–44, 96—98, Chapter 7).

The key difference between cost and risk lies with the fact that whereas a signal’s cost will be incurred when one delivers the signal, the risk as-sociated with the signal will not be realized until the other side rejects or even takes advantage of it. A signal’s risk is thus the additional loss that the signal’s sender will suffer if the signal is not reciprocated, multiplied by the probability that it will not be reciprocated.17


If the other side reciprocates one’s cooperative gesture, however, one can recuperate some of the gesture’s cost. Indeed, one can gain additional benefits from cooperation, such as saving precious resources by limiting arms race, increasing trade by reducing trade barriers, and increasing mu-tual trust, if both sides can cooperate with each other. One thus needs to take some risk of being cheated by others’ uncooperative behavior in order to get the benefit of being rewarded by others’ cooperative behavior. The key is to have a balance between the need to minimize the potential loss from others’ defection and the need to induce cooperation from others and reap the benefits from such cooperation. A state’s calculation of the risk of being penalized if the other side defects and the potential reward from cooperation if the other side reciprocates is a crucial calculation for a state that is considering a reassurance signal (Deutsch 1958, 265–269; Kydd 2005, 198–200).

The true cost of a reassurance signal is the private information of the signal’s sender. In contrast, just as the credibility of a signal of threat (in an ongoing conflict) is the private information of the signal’s receiver (Mercer 1996; Tang 2005), the credibility of a reassurance signal is the private information of the signal’s receiver. In other words, a reassurance attempt’s true cost is nonrelational, whereas its credibility is relational. Whether a reassurance attempt is credible can only be determined by its receiver.18

Moreover, just as the credibility of a signal for resolve in conflict con-sists of three factors (i.e., interest, capability, and reputation for resolve; see Tang 2005, 37–40), the credibility of a reassurance attempt also consists of more than one factor. While an extensive discussion on the credibility of a reassurance attempt will have to be provided elsewhere, suffice to say that the credibility of a reassurance attempt depends on its perceived cost, risk (if not reciprocated), potential gain (savings in resources, regardless of whether the other side reciprocates, plus benefits when the other side recip-rocates), and the perceived situational or external constraints faced by its initiator. Intuitively, the costlier and riskier the signals are, the more cred-ible they become, and the less they can be explained away by situational constraints (Jervis 1970, 19; Mitchell 1991; Larson 1997; Kydd 2006, Chapters 7 and 8; Montgomery 2006, 177). More formally, CA—the credibility of a reassurance signal—is determined by the equation below:

CA = f ( )the attempt’s cost + risk − the attempt’s gains,

all as perceived by the receiverthe situational constraint faced by the initiator

as perceived by the receiver


Two points become obvious from this equation.19 First, the credibility of a reassurance attempt depends only partly on the attempt’s true cost for its initiator, but more on the attempt’s cost, risk, gains, and situational constraints perceived by the receiver. Generally speaking, there will not be an exact correlation between the attempt’s cost perceived by the receiver and the attempt’s true cost for the initiator.

Second, while it is sound to infer that a reassurance attempt is suffi-ciently credible from its success (i.e., for a reassurance attempt to succeed in convincing the other side that one is benign, the other side must have deemed the attempt to be sufficiently credible), it is often misleading to infer whether a reassurance attempt is costly enough from its success (e.g., Montgomery 2007, 177). At the very least, inferring whether a reassurance attempt is sufficiently costly by its success neglects the possibility that the attempt’s receiver may be so aggressive or fearful that even a very costly reassurance attempt, short of total submission, may not be credible enough (Larson 1997, 243; see also Weinberger 2003; Kydd 2005, Chapter 4).

Determining Success and Failure of Reassurance

Judging the success and failure of a reassurance attempt or program is inherently a subjective exercise, and ultimately only decision-makers can make the judgment. This subsection provides a guide for making these judgments, with full knowledge that decision-makers almost inevitably fall into the trap of cognitive biases and errors that are common to human cog-nition when judging reassurance.

To understand the success and failure of a reassurance attempt, it is nec-essary to divide reassurance attempts into two general categories: those that do not require reciprocity and those that do. Obviously, it is more difficult to gauge the success or failure of a reassurance attempt that does not require reciprocity. When the receiver of a reassurance attempt simply accepts the reassurance attempt (usually in the form of conciliatory move or conces-sion) without having to reciprocate, the sender of a reassurance attempt cannot be sure whether the gesture was taken as a sign of benign intentions or as a sign of weak resolve. In contrast, it is easier to judge the success or failure of a reassurance attempt that does require reciprocity: an attempt is judged to be a success only if it has been reciprocated. Since most reas-surance attempts and programs do require reciprocity (not surprisingly), I focus on this type of reassurance attempts and programs. Apparently, reci-procity in reassurance must mean “diffuse reciprocity” in Robert Keohane’s (1986) terminology, because the receiver of a reassurance attempt may well reciprocate with a benign move that is different from the original reassur-ance attempt in both scope, issue domain, and magnitude.


A successful reassurance attempt is one that elicits a positive response from the receiver, whereas a failed reassurance attempt is one that does not. A successful reassurance attempt thus succeeds in reducing tension somewhat and moves a bilateral relationship one step toward a less conflic-tual and/or a more cooperative mode by building a bit more trust between two states. In contrast, a failed reassurance attempt may leave the bilateral relationship as it was, or worse, and will move the relationship to a less cooperative mode because the sender now deems the receiver as not so friendly, or even hostile. A failed reassurance attempt can thus “succeed” in revealing the other side’s less than friendly intentions! Of course, a reas-surance attempt can also elicit an ambiguous response, thus leading to an inconclusive reading of the receiver’s intentions.

Positive responses toward a reassurance attempt, from the least to the most positive, include (1) noting the positive direction of the attempt; (2) noting the positive direction of the attempt and emphasizing common interest and possibility of cooperation; (3) noting the positive direction of

Table 5.1 Types of response to a reassurance attempt

Positive Response Negative Response Ambiguous Response

Noting the positive direction of the attempt.

Ignoring the attempt. Noticing the attempt but without acknowledging the benign nature of the attempt.

Noting the positive direction of the attempt and emphasizing common interest and possibility of cooperation.

Rejecting or denouncing the attempt as a (wicked) ploy or public stunt.

Demanding costlier gestures from the sender, without noting benign nature of the attempt.

Noting the positive direction of the attempt but demanding costlier (thus more credible) signals from the sender, with promise to reciprocate to future costlier signals.

Rejecting or denouncing the attempt as a (wicked) ploy and demanding costlier signals from the sender without promising to reciprocate.

Demanding costlier gestures from the sender, without promising to reciprocate to future costlier signals.

Reciprocating the attempt.

Taking advantage of the attempt and demanding costlier signals from the sender without promising to reciprocate in the future.

Reciprocating with a costlier thus more reassuring gesture.


the attempt but demanding costlier (thus more credible) signals from the sender with a promise to reciprocate to future costlier signals; (4) recipro-cating the attempt; and (5) reciprocating with a costlier, thus more reas-suring, gesture. Negative responses, from the least to the most negative, include (1) ignoring the attempt; (2) denouncing the reassurance attempt as a (wicked) ploy and/or a public stunt; (3) rejecting the reassurance attempt as a (wicked) ploy and demanding costlier signals from the sender without promising to reciprocate; (4) taking advantage of the attempt and demanding more from the sender without promising to reciprocate in the future.20 Ambiguous responses include (1) noting the attempt without not-ing the benign nature of the attempt; (2) demanding costlier gestures from the sender without noting the benign nature of the attempt; (3) noting the attempt and demanding costlier gestures from the sender, without promis-ing to reciprocate to future costlier signals (table 5.1).21

A successful reassurance program is a program that eventually leads two states to overcome—and sometimes, though rarely, eliminate—some signif-icant amount of uncertainty about each other’s intentions and become more confident of each other’s benign intentions. Consequently, the two states will begin to trust each other more confidently and cooperate with each other more extensively, and thus enter a generally cooperative relationship.

Major Errors and Inadequacies

With few exceptions, existing literature on coping with uncertainty about others’ intentions through reassurance suffers from some crippling errors and inadequacies.

“Trust Must Proceed before Reassurance”

Obviously, trust facilitates cooperation whereas mistrust limits coopera-tion. Thus, the notion that “[extensive] trust must proceed before coop-eration” approaches conventional wisdom among IR theorists. “Trust . . . is usually a necessary (though not a sufficient) condition for states to coop-erate” (Larson 1997, 6; see also 32, 36). Similarly, “Cooperation requires a certain degree of trust between states,” and “cooperation is possible when the level of trust for the other exceeds a minimum trust threshold for each party” (Kydd 2005, 5, 9; emphasis in original).

Unfortunately, some have taken this notion that “[extensive] trust must proceed before cooperation” to mean that extensive trust must exist before initiating reassurance.22 This is the most fundamental error in our under-standing about reassurance and trust, and it has systematically biased IR scholars to exaggerate the difficulty of achieving international cooperation and stray away from the true causes behind the difficulty. Thus, Glaser


states that signaling one’s own benign intentions (i.e., reassurance) should occur only after one knows the other side is benign (Glaser 1994–1995, 67–68). Likewise, despite recognizing the inadequacy of Glaser’s formu-lation, Montgomery too insists that “uncertainty [over others’ intentions] frequently prevents states from communicating their motives [i.e., initi-ating reassurance]” (2006, 157; see also 162).23 Both authors imply that initiation of reassurance requires a significant amount of trust, that is, (almost) a certainty that the other side is benign.

After explicitly differentiating reassurance from piecemeal and exten-sive cooperation and emphasizing reassurance as a means of forging coop-eration by building trust, it becomes clear that while extensive cooperation requires a significant amount of trust ex ante (Larson 1997, 12, 21), piece-meal cooperation—usually achieved through reassurance—requires only marginal amount of trust ex ante.

Moreover, piecemeal cooperation is actually a means of building trust. Although significant amount of trust must exist for extensive cooperation to materialize, this significant amount of trust can only come from gradual cooperation earlier, which is usually achieved via reassurance (Larson 1997, 21).24 The relationship between trust and cooperation is thus a feedback relationship: trust is both a cause and an effect of cooperation (Gambetta 1988, 225; Hardin 1992, 513–516; Larson 1997, 10, 21; Yamagishi et al., 2005; Kydd 2005, Chapter 7). Between two benign states, cooperation-building via reassurance—usually embodied in a reassurance program—increases mutual trust and reduces mutual mistrust, thus moving the two states into a more cooperative relationship. This increase in mutual trust is possible only because reassurance facilitates cooperation and more (piece-meal) cooperation leads to more trust.

Most importantly, reassurance does not require extensive trust ex ante. Reassurance actually depends on the logic that only minimal trust ex ante is necessary before signaling reassurance (Kydd 2000b, 415). Put differently, uncertainty over others’ intention is actually a precondition for initiating reassurance (see section 4 for details). If extensive trust has to come before any reassurance signal, no reassurance is possible. This is simply because reassurance actually is driven by the desire to build trust and reduce mis-trust, and building trust via reassurance fundamentally depends on taking some risk in the possibility that the other side is untrustworthy. Without putt-ing some of one’s welfare into the other side’s hands even though one does not trust the other side, it is impossible to build trust between oneself and the other side.

Yamagishi and his colleagues (Yamagishi et al. 2005) have provided compelling experimental evidences that once trust and cooperation are separated by differentiating piecemeal cooperation from extensive


cooperation, and thus allow players to see the possible feedback relation-ship between them, it becomes clear that initial cooperation without trust or with only very limited amount of trust leads to more trust and more extensive cooperation.25 In the IR literature, both Larson (1997) and Kydd (2005, esp. Chapters 7 and 8) have provided compelling historical evi-dences that reassurance does not require extensive trust ex ante and piece-meal cooperation leads to greater trust.26 Moreover, statesmen understand that reassurance as an invitation to cooperation is a means for building trust rather than the other way around. Clearly, when Gorbachev initiated his reassurance program toward the United States and allies, he had little if any trust in Ronald Reagan (Larson 1997, Chapter 6; Collins 1998; Kydd 2005, Chapter 8). Likewise, when Southeast Asian states decided to engage China in the early 1990s, they had little trust in China (Collins 2000, Chapter 5).

Signaling Benign Intentions versus Gauging Others’ Intentions

The second major error, which is closely related to the first error, is that most authors have generally treated signaling one’s own benign inten-tions and gauging others’ intentions as two separate things and contended that one has to know that the other side is benign before proceeding to reassure.

For instance, Glaser (1994–1995, 67–70) identifies reassurance as the solution to the problem of another state’s uncertainty over one’s own (benign) intentions, thus implicitly assuming that one already knows the other state’s benign intentions when one initiates reassurance.27 Glaser’s formulation thus “makes reassurances an effect, as well as, a cause of reduced uncertainty” and reflects the logic of “one-sided uncertainty” (Montgomery 2006, 154, 161). Yet, one must be left to wonder how one came to know that the other side is benign in the first place.

Correctly noting the defect in Glaser’s formulation, Montgomery makes apparent the dual challenges that a state faces, “Not only must a signaling state endeavor to reveal its benign preferences [i.e., intentions]; it must also attempt to discover whether its adversary is a security seeker” (Montgomery 2006, 183). Unfortunately, Montgomery sees the dual chal-lenges as essentially intractable: how can reassurance be the solution to another state’s uncertainty over one’s own intentions when one cannot know the other state’s intentions first? (162), As a result, Montgomery ends up in essentially repeating Glaser’s mistakes (Tang 2007).

Yet, reassurance really does accomplish together the twin tasks of gaug-ing others’ intention and signaling one’s own benign intention. Signaling one’s own benign intentions and gauging the other side’s intentions are


inseparable. Indeed, if one wants to learn about another state’ intentions, then, other than observing its behavior toward a third party, one has to be benign and actively play the reassurance game, because only genuine reassurance attempts “help to distinguish trustworthy [i.e., benign] states from non-trustworthy [i.e., malign] ones” (Kydd 2005, 19; see also sec-tions 4 and 5).

Reassurance as Ends or Independent Strategy

The third major error on reassurance is to take reassurance as an end by itself or an independent strategy rather than only a means toward the greater end of reducing uncertainty over others’ intentions. In reality, reas-surance is a means toward the larger end of knowing another state’s inten-tions and then increasing cooperation if those intentions are benign (i.e., the other state is a fellow defensive realist state). Moreover, even a reassur-ance program usually cannot be an independent strategy; it is usually a part of a broader strategy called engagement (chapter 4).

This incorrect formulation has led some to deny the most critical ratio-nale for reassurance. If reassurance is the end rather than the means toward greater ends, then there is little rationale for it when the other side may take advantage of it. Thus, Montgomery charges that “although defen-sive realists have focused on how security seekers can demonstrate their preferences [for cooperation], they have placed less emphasis on explaining why they would do so in the first place” (Montgomery 2006, 154). Yet, defensive realism has always had a ready explanation for why states should choose reassurance. States have powerful incentives to reassure each other if they want to achieve cooperation, because cooperation is beneficial, and reassurance is necessary for moving toward cooperation.

Treating reassurance as an independent strategy also leads Montgomery to advance arguments that defy common sense. He asserts, “Because states are more secure when defense is strong, they have virtually no incentive to attempt reassuring gestures that might undermine that security in the hope of overcoming uncertainty” (Montgomery 2006, 166; emphasis added). But why would this be the case? Does realism not assume that states are strategic actors? If so, why would not (defensive realist) states actually take advantage of a window of opportunity in which defense is strong to build more cooperative relationships, so that they can feel more secure in the event offense becomes strong?28

In the end, Montgomery essentially recites offensive realism’s man-tra, which stipulates that states should and will generally favor compe-tition because cooperation (through reassurance, building trust, and then extensive cooperation) is so much riskier than competition: “This


constraint on what military reassurance can safely accomplish will fre-quently leave competition the preferred option, unless a state believes that its adversary is and will likely remain benign” (Montgomery 2006, 184; emphasis added). Obviously, according to his logic, states must always favor competition because certainty about the future is generally impos-sible under anarchy!

Fundamentally, Montgomery fails to grasp the fact that just as cooper-ation can be risky, so too can competition, because it may waste precious resources that can be directed to some other more pressing fronts (Glaser 1994–1995, 58–60).29 Once we admit that both cooperation-building and competition are possible means of seeking security because both are poten-tially risky and (thus) rewarding, it becomes clear that defensive realism has always had a ready explanation why states should choose reassurance. Because cooperation is beneficial and reassurance is necessary for building cooperation, states have powerful incentives to reassure each other if they desire cooperation (Kydd 2005, 19; see also Glaser 1994–1995, 57–60; Kydd 1997b).

And apparently, some wise defensive realist statesmen had known bet-ter. They had pursued reassurance and cooperation when they saw the benefits, necessities, and opportunities for cooperation. They understood that while being cheated may cost them something, running some risk in pursuing cooperation is a worthy price because foreclosing the opportu-nity of cooperation and engaging in an unnecessary arms race or conflict is also costly. For example, in the early days of the cold war, both Henry L. Stimson and Dean Acheson supported some type of international control of atomic weapons to avoid the looming arms race between the United States and the former Soviet Union, and the rationales behind their stand were almost identical to the logic outlined here. Both Acheson and Stimson acknowledged that seeking cooperation with the Soviet Union involved some risk, but believed the risk was worth taking, as Acheson observed perceptually: “The advantage of being ahead in such a race is nothing compared with not having the race” (Quoted in Kydd 2005, 98).

“Cooperation or Reassurance for Inducing Benign Intentions”

The fourth major error is to frame the relationship between intentions and cooperation, and, by implication, between intentions and reassurance, incorrectly. For instance, David Edelstein (2002, 13) argues that “cooper-ation can increase the probability of benign intentions if it makes a state more secure, and thereby, mutes any expansionist ambitions,” and that “by demonstrating one’s own benign intentions [i.e., reassurance], a gov-ernment hopes to convince another state that it can safely adopt similarly


benign intentions.” For Edelstein, cooperation or reassurance is a means for inducing benign intentions (from an offensive realist state). This for-mulation is again misleading.30

As noted above, although cooperation-building is a means toward secu-rity when both states are benign (Glaser 1994–1995; Jervis 1999), it is not a means for inducing benign intentions. Cooperation can merely reinforce mutual benign intentions, but cannot induce benign intentions. Rather, reassurance is a means for inducing benign behaviors from a benign state that is fearful. By reducing another state’s fear about one’s own intentions, the other state can be expected to reciprocate with moderate behaviors and cooperation can then be forged.

By no means can one hope to convince a malign state to change into a benign state by merely signaling one’s own benign intentions or by cooper-ating. In fact, when facing an offensive realist state, a strategy of only coop-eration is just appeasement and will almost surely embolden an offensive realist state to pursue its aggressive goals rather than turn it into a defensive realist state. To change another state’s intentions from malign to benign requires more than cooperation—it requires the more complex strategy of engagement (chapter 4).

Edelstein’s mistaken formulation doomed cooperation—usually achieved via reassurance—to be dangerous under most circumstances. Because cooperation is not supposed to induce benign intentions, it is not surprising that Edelstein found that “cooperation did not induce benign intentions in any of the cases, and, in at least the case of interwar Germany, cooperation was tragically misguided.” Edelstein (2002, 38, see also 15–17, 38–40) thus concludes that cooperation is “often a poor stra-tegic choice.”31

A closer look shows that Edelstein’s whole theoretical and empirical enterprise is flawed. Theoretically, Edelstein starts from a questionable assumption: “The most common theory that governments hold is that malign intentions are generated by insecurity” (Edelstein 2002, 13; emphasis added). As Osgood (1962, 18–36) and Jervis (1976, esp. 67–76, 349–355) noted long ago and the enormous literature of attribution theory in social psy-chology suggests, however, the exact opposite is true: states tend to believe that others’ malign intentions are caused by others’ inherent evil nature.

Edelstein’s reading of the coming of the cold war is also flawed and often self-contradictory.32 He claims that “uncertainty about Soviet’s intentions [between later 1945 and early 1946] led the United States to pursue competitive strategy” (34). Yet, most analysts will argue that between September 1945 and March 1946, the United States was quickly concluding that the Soviet Union under Stalin was a genuine expansionist power (e.g., Gaddis 2005, Chapter 2; Kydd 2005, Chapter 3; Weinberger


2003, 110–113). In fact, Edelstein (2002, 33–34, 37) admitted as much: “In later 1945 and early 1946, the United States concluded from Soviet behavior in Iran, Turkey, and Germany that allaying Soviet security con-cern was beyond the means Washington was willing to consider . . . As a consequence, the United States opted for competition rather than coop-eration.” Thus, the United States hedged (i.e., it mixed competition with cooperation) only when it still had some uncertainty about Stalin’s intentions. As the United States became more certain about Stalin’s expansionism, it took a progressively firmer stand against the Soviet Union and the hedging element in the strategy of the United States decreased steadily. As Edelstein (2002, 12) contended earlier, states hedge their bets when there is uncer-tainty: More the uncertainty, more the hedging; less the uncertainty, less the hedging.

Overreliance on Offense-Defense Balance

Many authors have heavily relied on the offense-defense theory (ODT)—or more specifically, the differentiation of military postures (and perhaps some weapons) into offensive and defensive types and the offense-defense bal-ance (ODB)—to operationalize their arguments about the conditions for signaling and reassurance (Jervis 1978; Glaser 1992; 1994; Montgomery 2006). Moreover, ODB has received far more attention than the differ-entiation of military postures. Unfortunately, whereas differentiation is a valid component of ODT (discussed later), ODB is not (Tang 2010b).

ODB has two different versions, an objective one and a subjective one (i.e., the balance as perceived by states), and the two versions have very dif-ferent implications for understanding international politics. The objective offense-defense balance influences only the outcome of war, and only the sub-jective offense-defense balance influences states’ calculation of war and peace, including reassurance.33 Because many fail to recognize the two versions of the balance and their different implications, they have simply conflated the two versions of the balance or anchored their discussion of states’ cal-culation on reassurance upon the objective balance when in fact they have to anchor their discussion upon the subjective balance (e.g., Glaser 1994–1995, 61–67; Montgomery 2006, 163–167).

As I show in greater detail elsewhere (Tang 2010b), objective ODB is almost impossible to measure and operationalize, either with the technology- only approach (i.e., a narrow approach) or with the technology- plus approach (i.e., a broad approach), unless we are dealing with extreme situations. Discussions centered upon subjective ODB may be on to some-thing, but the theoretical enterprise centered upon the subjective balance faces two fundamental problems.


The first problem is that proponents of the subjective balance have yet to provide convincing evidence that statesmen have made their decisions on war and peace based on their judgment of the balance. So far, the only real case that supports the impact of the subjective balance on decision-makers has been WWI in Europe.34 The second problem is that even if we assume that statesmen do often make their decisions on war based on their judgment of the (objective) balance, proponents of the subjective balance have yet to really touch upon the more fundamental aspect of the puz-zle. They have failed to grasp that the subjective balance, especially states’ belief in offensive advantage, is not an independent variable that is ideally made for explaining the cause and timing of war but really a dependent variable that begs for a theory (Finel 1999, 188; Gray 1993, 39; Shimshoni 1991, 197–201).

Because many have relied on ODB to operationalize their arguments about the conditions for initiating reassurance, much of the existing lit-erature on reassurance through military postures has a distinct “apoliti-cal” flavor, thus leaving other more important factors out of the discussion and preventing an adequate understanding of the crucial problem. Worse, because (objective) ODB is on shaky ground, much of the discussion has inevitably come to rest on sand.

Offense and Defense Capabilities, Total Power, and Reassurance

Despite having focused mostly on reassurance attempts within the mili-tary domain, existing literature has also gotten many critical issues about reassurance within the military domain incorrect.

First, while existing literature has regarded the differentiation of weap-ons/military capabilities and military postures into offensive and defensive types as crucial for reassurance since Jervis’s seminal contribution (Jervis 1978), it has often focused solely on weapons. Yet, because very few weap-ons or weapon systems (e.g., fixed fortress, aircraft carrier) can be unam-biguously classified as offensive or defensive and states deploy an arsenal of weapon systems rather than just some weapons, we need to focus on military capabilities and postures when trying to understand reassurance within the military domain (Tang 2010b).35

Second, existing literature has often treated differentiation as a system-wide variable. Yet, under most circumstances, differentiation of a state’s military posture can only be achieved by measuring it against other states’ military postures. Hence, only a dyadic approach can work for differentia-tion of military postures, consistent with the notion that we need to focus on relative military capabilities rather than relative (total) power when discussing reassurance with military posture (Glaser 1994–1995, 60–64,


68–70). For instance, when examined in isolation, a navy is undoubtedly an instrument for offensive purpose because it is only useful for projecting forces away from a state’s homeland. Yet, in today’s age in which many states possess a navy, only a navy that centers upon a fleet of aircraft carri-ers is capable of performing serious offensive campaign away from home. As such, viewed in any dyadic situation, the United States navy has the most offensive posture. Finally, the United States military as a whole, with its heavy emphasis on projecting power to every corner of the world, maintains the most offensive posture in the world, viewed in any dyadic situation.

Third, defensive realism has yet to state explicitly that reassurance is still possible even when differentiation is difficult or impossible. In princi-ple, states can simply reduce all types of arms and troops along their bor-der to reassure each other, without rigorously differentiating offensive and defensive weapons. Two prominent cases in recent history are the gradual yet sustained improvement of the Russia-China and the Sino-Vietnamese relationship in the past two decades (Gu and Womack 2000; Dittmer 2004). Of course, because reassurance becomes very difficult and poten-tially very risky when differentiation is difficult or impossible, states may well decline to initiate reassurance under the circumstances. Nonetheless, because military postures can almost always be differentiated within a dyad, states can almost always signal benign intentions with their military postures if they choose to.

Finally, existing literature tends to conflate offensive and defensive capabilities and focus on total relative power rather than on relative mil-itary capabilities, and it often does not apply the logic of the differentia-tion consistently during the discussion. For instance, after correctly noting that a defensive realist state must “go beyond tokens, and make conces-sions weighty enough so that a state contemplating attacking or coercion would be unwilling to make them,” Kydd asserts that “the concessions are inherently risky and would not carry convictions if they were not” (Kydd 1997b, 144–145, n81). This statement is imprecise and thus misleading, because it does not explicitly state that the concessions should be risky for what purposes, and one is left with the impression that any concession is risky. More seriously, after citing Kydd’s formulation above, Montgomery (2006, 158; see also 153, 158–160, 163–164, 184) then asserts that “the primary way a state can reveal benign motives is by taking actions sub-stantial enough to decrease its ability to defeat an adversary in war, if one were to occur.”36 As a result, Montgomery dooms reassurance to be dan-gerous under most circumstances. Finally, Copeland (2000, 40) asserted that “conciliatory reassurance may reduce the probability of major war breaking out as a result of an inadvertent spiral. But by sacrificing relative


power in the process, it can lower a state’s likelihood of winning any war that does occur.” Apparently, Copeland talks about total (relative) power and does not differentiate between the power to win an offensive war and the power to win a defensive war. His statement has to be qualified: a state can actually win a defensive war without being able to win an offensive war, and vice versa.37

All these statements are underspecified and thus misleading. The notion that a defensive realist state can reassure an adversary only by tak-ing actions that decrease it ability to defeat the adversary in the event of a conflict holds only when military capabilities cannot be differentiated. When military capabilities can be differentiated, reduction in unnecessary offensive capabilities may actually enhance the security of a defensive real-ist state because this allows it to concentrate more resources on defense.

Because relative power—when undifferentiated—is an important pillar for security, states will be very reluctant to sacrifice undifferentiated rel-ative power for reassurance. Existing literature thus underemphasizes the critical point that while states may be unwilling to sacrifice undifferenti-ated relative power for reassurance, they may be willing to sacrifice some offensive military capabilities for reassurance. As a result, existing litera-ture has underestimated the feasibility of reassurance. When discussing reassurance attempts in military aspects, we have to differentiate relative (total) power from (relative) military capabilities/postures and focus on the latter rather than the former (Glaser 1994–1995, 60–64, 68–70).38

Forms of Reassurance Attempt

A state can deploy three general types of reassurance attempts: words, non-military deeds, and military deeds (see table 5.2 for a summary). A reassur-ance program normally combines all three types, although a receiver may pay more attention and confer more credibility to reassurance attempts within the last two types.

Reassurance with Words: Some Words are not Costless

When signaling resolve in conflict, only behaviors that carry some material cost can convey some credibility, and words (i.e., rhetoric) generally carry little weight because words generally carry little material cost (Schelling 1966, 150; Fearon 1994). When signaling benign intentions, however, words can also carry some—although not a lot—credibility because they can incur some symbolic cost upon and pose some risk for the sender, even if they incur little material cost (Jervis 1970, 35–37; Larson 1997, 28; Stein 1991, 432). Moreover, while a reassurance program with words alone


surely cannot move a relationship to a very cooperative mode, words are an integral part of any reassurance game because they often not only provide the context for the other side to interpret one’s deeds but also set the right climate for reassuring with deeds.39

From the least credible to the most credible, there are at least five gen-eral forms of reassurance with words. The first is to stop labeling the other side as one’s archenemy, usually by toning down one’s accusations and/or propaganda war against the other side. This form is perhaps a necessary

Table 5.2 Forms of reassurance attempt

Reassurance with Words Reassurance with Deeds (nonmilitary)

Reassurance with Deeds (military)

Stopping calling the other side as one’s archenemy and stopping the propaganda war against the other side.

Granting some small concessions that were requested by the other side.

Stopping supporting insurgencies against the other side.

De-emphasizing the seemingly irreconcilable conflict of interest and emphasizing “misunderstandings” that can be resolved or patched up by dialogue, consultation, and negotiation.

Reversing one’s previous policies that were designed to punish the other side.

Halting the deployment of troops and the deployment/development of weapons.

Calling for calmness, consultation, dialogue, and negotiation to resolve ongoing disputes, especially when a dispute seems to heat up and nationalism in both sides is on the rise.

Initiating some exchange and cooperation or broadening the scope of existing exchange and cooperation.

Adopting a defensive doctrine/posture and forsaking some unnecessary offensive capabilities.

Breaking a taboo, going against the domestic consensus, and reversing one’s confrontational rhetoric and adopting moderate rhetoric.

Bold and irreversible political moves that signal benign intentions.

Reduction of one’s overall military capabilities, especially offensive capabilities.

Public pledges or promises to cooperate and help when the other side is in distress or misfortune.

Extending a helping hand to others when others are in distress.

Joining and creating security regimes that limit competition and facilitate cooperation.

Abiding by existing rules and norms or creating new rules and norms that regularize and facilitate cooperation.

Helping the other side militarily when it is in needy time.


first step between two former foes that have had a long history of confron-tation. Thus, when Nixon was preparing his opening speech to China in 1970, he began to address China by its official name, People’s Republic of China, rather than Red China. Similarly, when the former Soviet Union and China began their long journey toward reconciliation in 1984–1985, they began to tone down their rhetoric against each other. Likewise, when South Korea under the then president Kim Dae-jung launched a reconcil-iation policy (i.e., “Sunshine Policy”) toward North Korea, South Korea began to address North Korea by its official name (Democratic People’s Republic of Korea, or DPRK) and halted its long-running propaganda war against North Korea.

The second form is to de-emphasize conflicts of interest and emphasize “misunderstandings” that can be resolved or patched up by dialogue, con-sultation, and negotiation. For example, when Ronald Reagan was ready to engineer some kind of de-escalation with the Soviet Union in 1983–1984, he began to emphasize “misunderstandings” between the two countries that could be cleared up, and in addition he quietly stopped calling the Soviet Union the “Evil Empire” (Larson 1997, Chapter 6).

The third form is to call for calmness, consultation, dialogue, and nego-tiation to resolve ongoing disputes and prevent a dispute from disrupting the overall bilateral relationship, especially when a dispute or conflict of interest seems to heat up and nationalism in both sides is on the rise. This might be the most typical form of reassurance with words, and examples are everywhere. Thus, when the dispute between China and South Korea on the historiography of the ancient Koguryo/Gaogouli kingdom seemed to be erupting into a full crisis in June 2003, China sent one of its vice for-eign minister to reassure South Korea and convey the message that Seoul and Beijing should try to work together to limit the dispute and prevent it from affecting the overall bilateral relationship.

The fourth form is rhetoric that breaks a taboo, goes against one’s domestic consensus (i.e., swims against the current), or simply reverses one’s old confrontational rhetoric. Two examples from the cold war years fall into this category. Khrushchev denounced the slogan of “peaceful co-existence” during his power struggle versus Beria and Malenkov but then picked it up after consolidating power in 1953. Similarly, John F. Kennedy’s June 1963 speech at the American University called for a criti-cal examination of the role of the United States in the cold war.40 A more recent example in this category is President George W. Bush’s shift from labeling China as “a strategic competitor” to “a stakeholder” in 2005. This form of reassurance via words carries some credibility because it entails some real audience cost for the leader. Khrushchev’s “peaceful co-exis-tence” was bitterly attacked by the more radical China as “revisionism,”


whereas Kennedy’s 1963 speech was attacked as “a dreadful mistake” by Congressional republicans (Larson 1997, 28), and Bush’s reversal of course amounted to a slap on his own face in front of his domestic audience.

The fifth form is public rhetoric and verbal promises to cooperate and help. Such publicly voiced rhetoric and verbal promises may become extremely costly to be recanted diplomatically. More importantly, this form of reassurance via words often implicitly and explicitly requires deeds to match the words uttered.41 As such, this form of reassurance via words car-ries more credibility than the other forms mentioned above. Thus, when a state wants to demonstrate its goodwill toward another state, the former usually pledges to help when the later suffers from a natural disaster or some other misfortune. A classic example in this category is China’s pub-lic pledge to not devalue its currency during the height of the 1997 Asian financial crisis. More recent examples include other countries’ expressing solidarity and pledging to support the United States after the terrorist attack on September 11, 2001 and Pakistan’s pledge to aid India’s investi-gation of the terrorist attack in Mumbai on November 26, 2008 when it became clear that some of the attackers were from Pakistan.

Reassurance with Deeds: Nonmilitary

Reassurance with deeds in nonmilitary issues involves specific moves or behaviors in economic or political areas. While these behaviors may have little direct implications for military affairs, they can still convey benign intentions. Indeed, between two states with no active military confron-tation, reassurance often involves reassurance with deeds in nonmilitary issues. Even between adversaries in active confrontation, reassurance attempts with deeds in nonmilitary issues are still useful—and often indis-pensable—for improving the atmosphere and paving the way for reassur-ance with military deeds later.

The first form is to grant some small concessions required by the other side. Thus, when President Reagan requested the Soviet Union to grant exit visas for seven Pentecostal Christians who had been living in the U.S. embassy in Moscow since 1978, the Soviet Union under Andropov agreed, even though it deemed the request “odd and suspicious” (Larson 1997, 192–193).

The second form is to reverse one’s previous policies that were designed to punish the other side. Perhaps the most common gesture in this cat-egory is the lifting of a trade and tourism embargo that one had earlier imposed upon the other side. When Britain wanted to repair its relations with Argentina after the Falklands/Malvinas war, it lifted trade restric-tions on Argentina (Mitchell 1991). Similarly, when the Bush administra-tion decided to reverse its previous hard-line policy toward North Korea, it


began to relax its economic embargo on North Korea and promised to lift more restrictions (e.g., taking North Korea off from the list of states that sponsor terrorism). Resumption of summit meetings between two states that had been cancelled or postponed due to some ongoing disputes is the more symbolic gesture in this category.

The third form is to initiate and broaden some (new) mutual exchanges and cooperation or deepen the existing ones. Under the rubric of cultural and sports exchange, the most famous example might have been the “ping-pong diplomacy” between the United States and China in 1971, and the Boston Symphony Orchestra’s visit to the Soviet Union in 1956. Russia’s and China’s joint launching of the “Year of China” in Russia and the “Year of Russia” in China to cement their strategic partnership is a more ambitious example in this category. The launching of a series of “strategic economic dialogue” (SED) between the United States and China after President Bush did away with the label of “strategic competitor” for China and relabeled China as a “stakeholder” in 2006 is a more recent example.

The fourth form is dramatic gestures of reconciliation. A classic exam-ple is (West) German Chancellor Willy Brant’s dramatic kneeling in 1970 after laying a wreath at the memorial to those who were killed during the 1943 uprising in Warsaw, a gesture that went far beyond the diplomatic norm. An equally dramatic gesture of this kind was Egyptian President Sadat’s visit to Israel in 1977. Both moves were taken by their intended receivers as extremely credible moves of benign intentions, partly because both are essentially irreversible (Maoz and Felsenthal 1987; Stein 1991, 441–444).

The fifth form is to extend a helping hand when others are in dis-tress and facing misfortune. Thus, countries (e.g., Japan and China) that extended help to the Southeast Asian countries and South Korea during the 1997 financial crisis were showered with appreciation by the Southeast Asian countries and South Korea (whereas the United States suffered a sig-nificant loss of goodwill because it forbade the IMF to rescue these coun-tries with better conditions). More recent examples include Australia’s, China’s, Japan’s, and the United States’ rapid and large-scale response to Indonesia’s tsunami in 2004 and Japan’s relief effort in the quake zone after the devastating earthquake in Sichuan province in China.

The sixth form is to submit to or create rules and norms that codify cer-tain cooperative behaviors and organizations that enforce those cooperative behaviors in nonmilitary affairs (Nye and Keohane 1977; Keohane 1984; Stein 1991, 444–449).42 A recent example is the “Chiang Mai Initiatives” that promise multilateral rescue for countries hit by a future financial cri-sis, launched by China, Japan, South Korea, and the ten Southeast Asian countries in the aftermath of the 1997 Asian financial crisis.


Signaling with Deeds: Military

States can also deploy several measures in military affairs to signal its benign intentions.43 The first form, especially useful for two former foes, is to stop supporting insurgencies on the other side. This was one of the first moves that China undertook when it initiated its long sustained peace initiative toward the Southeast Asian countries. While China rigorously supported insurgencies in many Southeast Asian countries based on com-munist ideologies from 1960 to the 1970s, China cut off this kind of sup-port completely after 1978.44

The second form is to halt deployment of troops and deployment/devel-opment of certain weaponries. This was Gorbachev’s initial gesture toward the West in 1985. He imposed a moratorium on nuclear testing and halted the deployment of SS-20 intermediate range ballistic missiles (IRBM). Although this gesture was not reciprocated by the United States, it created a favorable climate for more credible measures later (Collins 1998, 205; Goldstein and Freeman 1990, 116).

The third form is to adopt a more defensive military posture. This can be a redeployment of troops from the border to a more defensive position or a general reduction of troops along the border. In the earlier days of their rapprochement, Soviet Union and China not only redeployed their troops to a more defensive posture but also substantially reduced the number of troops along their border. The rapprochement between India and China from 1989 to 1997 had involved similar measures.

The fourth form is to reduce one’s overall military capabilities, espe-cially offensive capabilities, without necessarily jeopardizing one’s capa-bilities for defense and deterrence. Troop reduction is perhaps the most common form within this category. Both Khrushchev’s reducing Soviet ground forces by 1.2 million in 1956 (in addition to the cuts before it) and Gorbachev’s reducing Soviet ground forces by 500,000 in two years and pulling many troops and equipments out of Eastern Europe in 1988 fall into this category (Larson 1997, 223–224).

The fifth form is to join together in creating security institutions and regimes that limit future security competitions and facilitate future secu-rity cooperation (Jervis 1982; Stein 1991; Weinberger 2003). All formal treaties that codify a resolution of conflict or conflict of interest belong to this type, so do regimes that are designed to limit security competition and regimes that facilitate regular military diplomacies. The Antiballistic Missile treaty between the Soviet Union and the United States in 1972 and the declaration of the Code of the Conduct of Parties in the South China Sea, between China and Southeast Asian states, in 2002 belong to this category.


The sixth form of reassurance with military deeds is to extend help to the other side when it is in needy time. Such an act may eventually cul-minate in an alliance. A prominent example that did result in an alliance was the “entente cordial” between Britain and France in 1904 in response to Germany’s growing assertiveness. A recent example that has no real pos-sibility in becoming an alliance any time soon, but may build more trust between the two sides, is China’s cooperation with the United States (and its allies) on the looming North Korea nuclear weapon crisis in a time when the United States is tied up in the Middle East and Afghanistan.

Nonevents as Reassurance Signal

A particular type of reassurance signal, usually as deeds, is nonevents. Because nonevents are “dogs-that-did-not-bark,” this type of reassurance signal has received virtually no attention in the existing literature. Yet, some nonevents actually carry powerful reassuring messages and thus deserve special attention.

For instance, in light of a powerful state’s military capabilities or total power, certain actions that are not taken by the state should be consid-ered as reassuring. If a state is vastly more powerful than its neighbor-ing states and yet it has chosen to behave amicably toward its neighbors, then the state is exercising some self-restraint and should be considered to be more likely to be benign, even if there are external factors in facili-tating its benign behaviors (Kydd 1997, 143). By the same token, a more powerful state’s demonstrated willingness to make concessions and reach compromise with its weaker neighbors should also be considered as reas-suring. Thus, in light of China’s growing power in the past three decades, its general reluctance to use force and its willingness to settle territorial disputes with its neighbors amicably and to generally cooperate with them has been recognized by its neighbors as reassuring (Collins 2000, Chapter 5; Shambaugh 2004–2005; Fravel 2008; Glosny 2006). A simple coun-terfactual exercise will substantiate this point. Just imagine what kind of conclusion external observers would have drawn if China did not settle any of its disputes, or worse, if it often resorted to land-grabbing using force.

Similarly, if a state indeed possesses the technological and financial capability to develop certain offensive capabilities, then its refraining from developing those capabilities is at least somewhat reassuring. Thus, in light of Japan’s technological prowess, Japan’s policy of abstaining from devel-oping nuclear weapons is reassuring to its neighbors (Midford 2002). The same can be said about Germany’s decision to stay away from nuclear weap-ons. Again, a simple counterfactual exercise will drive this point home. Just imagine the possible uproar among Germany’s and Japan’s neighbors


(e.g., Poland, France, China, and the two Koreas) if these two countries had gone nuclear.

Another form of nonevent that conveys a reassuring message is refrain-ing from preying on a state when it is in a period of apparent vulnera-bility (Jervis 1970, 20; 1976, 43).45 For instance, one can refrain from exacerbating the other side’s domestic instability. Here, refraining from expanding war aims when one is winning can be understood as a specific form of nonevent involving unilateral self-restraint: it at least signals one’s limited goals.46 A prominent example was the U.S. decision of not expand-ing the Korean War into China’s territory after China had intervened in the war. Another example was China’s unilateral ceasefire and then with-drawal back to the ex ante line of actual control (LOC) after defeating the Indian Army in the 1962 India-China border war. At the very least, these measures of restraint signaled a state’s limited goals in conflict, even if logistical concern might have been an important or even major cause for these self-restraints. Again, a simple counterfactual exercise will prove the point. Just imagine how much animosity China would have held against the United States if the latter had expanded the war into China, and how much animosity India would have held against China if China had actu-ally pressed ahead with its offensive drive or held on to some of the dis-puted territory after the ceasefire.

Unfortunately, while such reassurance attempts (from an active adver-sary) signal some benign intentions, or at least limited goals, the (vulnera-ble) receiving state is most unlikely to embrace it. When a state is defeated or faces defeat in a conflict, it is unlikely to take its opponent’s show of magnanimity as too soothing. Moreover, we tend to ignore nonevents because nonevents are “dogs-that-did-not-bark,” and we generally only carryout counterfactual exercises with nonevents.

Initiating, Reciprocating, and Succeeding: Conditions and Factors

Although conditions and factors for initiating, reciprocating, and succeed-ing with reassurance are linked and somewhat similar, they warrant sepa-rate treatment because they are sufficiently different.

Initiating Reassurance

An offensive realist state is generally uninterested in knowing other states’ intentions. It assumes others to be inherently aggressive and merely wants to know others’ capability and resolve in resisting its aggressive goals, and other states can appear friendly to it only by appeasing it or bandwagoning


with it. An offensive realist state thus sees little value in trying to cope with uncertainty through reassurance: “There is little room in offensive real-ism for a strategy of reassurance” (Montgomery 2006, 155; see also Kydd 2005, 183; Mearsheimer 2006).

Thus, the first and foremost condition for initiating genuine reassur-ance is that a state must be genuinely benign or a defensive realist state. This is simply because only a defensive realist state will be genuinely inter-ested in knowing others’ intentions and thus exploring the possibility of extensive cooperation.

The second condition is that some uncertainty about others’ intentions is sine qua non for initiating reassurance. Without uncertainty about oth-ers’ intentions, there is little need for reassurance (see section 3). Of course, because uncertainty about others’ present and future intentions cannot be completely eliminated—intentions can change, after all—reassurance is constantly required to sustain a cooperative relationship between two be-nign states. Hence, increased extensive cooperation can be understood as a reassurance gesture. Sustaining extensive cooperation continues to reas-sure both sides of each other’s goodwill.

The third condition for initiating reassurance is that a benign state must not be an extremely fearful (i.e., very insecure) state. Because initi-ating reassurance normally requires a state to take a conciliatory first step, an extremely fearful state—even if it is a genuine benign state—will be quite unlikely to initiate reassurance, all else being equal. Put differently, the more secure a state feels, the more likely it will initiate reassurance, because it will perceive the likelihood of having a disastrous outcome from another state’s defection to be lower (Kydd 2005, 183).47 Obviously, this logic brings all the physical and psychological regulators of the severity of the security dilemma or spiral (or fear) into the picture (Jervis 1978).48

Clearly, two states are far more likely to initiate reassurance when they face a common adversary, because there is an instrumental need to do so. Having a common ally is also conducive to reassurance (although perhaps less so than having a common adversary), because the common ally can induce and even pressure two states to move toward a more cooperative relationship for the sake of the alliance. The reconciliation between France and Germany after WWII thus had the double blessing of having a com-mon threat in the Soviet Union and a common (and powerful) ally in the United States; whereas the “entente cordial” between Britain and France in 1904 was mostly facilitated by the two countries’ increasing wariness of Germany’s assertiveness.

Of course, while the presence of common adversaries or allies is con-ducive to the initiation of reassurance, it is insufficient for a reassurance program to succeed, at least not completely. Thus, although South Korea


and Japan (during the entire cold war) and China and Japan (from 1972 to 1990) also had a common adversary in the Soviet Union and a common ally in the United States, reconciliation within the two dyads remains far from being a complete success.

Reciprocating Reassurance

A malignant state will be unwilling to reciprocate to genuinely costly con-cessions. Instead, it is likely to dismiss the attempt or to accept the attempt without reciprocating (i.e., take advantage of the concessions offered). Thus, an important condition for reciprocating reassurance is that the receiver is a defensive realist state.

Things actually become a bit more complex when the receiver is a defen-sive realist state. A defensive realist state that is extremely fearful is unlikely to reciprocate to the initial reassurance attempt offered by a benign state, mostly due to its fear that the initiator may be a (smart) malign state that is playing tricks. More likely than not, it will dismiss the initial reassurance attempt offered by the initiator as inadequate and demand more credible reassuring measures. After all, the initial reassurance attempt is usually a not too costly gesture. While a defensive realist state that is extremely fear-ful may dismiss the initial benign gesture from the sender as inadequate, it is less likely to take advantage of the benign gesture than an offensive realist state.

When the receiver is a not so fearful defensive realist state, it is more likely to reciprocate to the sender’s reassurance attempt. Although a reas-surance program is by no means guaranteed to succeed or succeed easily even in this situation, it is more likely to succeed than the situation with a very fearful defensive realist state.

Overall, factors that influence initiation of reassurance also influence the reciprocating of reassurance, although perhaps less significantly, sim-ply because the other side has made the first conciliatory move.

Succeeding with Reassurance

Apparently, factors that facilitate initiation and reciprocating of reassurance also make a reassurance attempt or program more likely to succeed. Yet, because it takes two to make a reassurance attempt or program succeed, whether a reassurance attempt or program succeeds depends on far more factors than either the initiation or the reciprocating of reassurance.49

Two states can build extensive cooperation only through a series of reassurance attempts or a reassurance program, and one of them has to take the first step. Thus, the first two necessary conditions for a successful


reassurance program are (1) that both states are defensive realist states and (2) that one of them feels secure enough and/or has enough incentives to initiate reassurance (e.g., facing a pressing threat from a third state). Hence, when the receiver is an offensive realist state, a reassurance attempt is unlikely to succeed and the program is almost sure to fail. In other words, an obvious explanation for the failure of a reassurance program is that its receiver is a very fearful state, or worse, an offensive realist state. Obviously, this possibility partly explains why even a well crafted reassur-ance program may not succeed (Larson 1997, 243; see also Weinberger 2003; Kydd 2005, Chapter 4).50

Because even a benign receiver may not reciprocate to the sender’s ini-tial reassurance attempts and yet demand more credible signals from the sender without promising to reciprocate, the receiver may well prevent a reassurance program from succeeding. When several of its initial reassur-ance attempts are dismissed as inadequate and not reciprocated, the sender may well conclude that the receiver is a malign state, thus deems it too risky to press ahead with more reassurance attempts within a reassurance program. In other words, a defensive realist state that initiates the reassur-ance may stop reassurance well short of the threshold level of reassurance that is necessary for moving the relationship into a more cooperative mode, for fear of being exploited if it presses ahead with costlier reassurance ges-tures. At this juncture, the receiver may also conclude that the sender is not a genuine benign state because the sender has refused to deploy more cred-ible signals. Here lies the real cause of the tragedy of the security dilemma. Although two states are desperately trying to avoid an arms race and a con-frontation and are eager for a more cooperative relationship, they cannot reach that goal because the deep fear held by one side may well prevent the (deep) security dilemma between them from being alleviated.

Thus, a third necessary condition for a successful reassurance program is a determined and not so fearful defensive realist state. Unless such a state exists, successful reassurance between two benign states is difficult to materialize.

Reassurance: Core Logic and Practical Guide

The core logic of reassurance can be stated briefly as follows. Strategically, reassurance is driven by a defensive realist state’s desire to forge a more cooperative relationship with another state. Operationally, reassurance is a means for gauging others’ intentions by signaling one’s benign intentions and then reading others’ intentions from their reactions toward one’s signal of benign intentions, and it critically depends on incurring some cost and taking some risk in the possibility that the other side may not reciprocate or even take advantage of one’s benign gesture.


Reducing uncertainty over others’ intentions can go both ways. When one ends up in discovering that the receiver is actually malignant, one should then adopt a firmer approach toward the other side. When one ends up in discovering that the other side is also benign, one should then aim for more extensive cooperation to cement the relationship. When facing a fellow defensive realist state, the logic is no longer for mere reassurance. Instead, two defensive realist states have every incentive to move beyond mere reassurance and onto the higher plateau of more extensive coopera-tion, because doing so enhances their security and welfare.

From this core logic, a practical guide to reassurance can be offered.51 One starts by issuing a reassurance signal (i.e., extending a genuine invi-tation to cooperate). When it comes to designing specific reassurance attempts, one has to strike a delicate balance. A reassurance attempt must be costly and risky enough, which a malign state would not consider mak-ing, but not too costly that being cheated will cost one dearly.

When it comes to actual signaling, five simple principles should be observed. First, words must come first. State clearly that one intends to move toward a more cooperative relationship with the other side. Second, speak with one voice. Speaking with one voice conveys seriousness of one’s message and consensus within one’s government, thus increasing the chance that the other side will be impressed. In contrast, dissenting voices that contradict benign intentions are noises, making interpretation of one’s reassurance attempts difficult.52 Third, state the specific steps one intends to take before one actually implements the steps. Fourth, after announcing one’s specific steps, match words with deeds even if the other side is not initially impressed by one’s pronouncement.53 Fifth, after the deeds are initiated, state clearly that one expects reciprocity from the other side, but avoid issuing specific demands.

After the reassurance attempt, one then gauges the other side’s true intentions from its reactions to the attempt. If the other side reciprocates, then it is more likely to be a fellow benign state. In contrast, if the other state responds by taking advantage of one’s goodwill, then it is more likely to be a malign state.

If the other state simply dismisses one’s reassurance attempt as insuf-ficient without taking advantage of it and demands for a more credible gesture, then it can be either an extremely fearful benign state or a very strategic genuine malign state. Because one’s initial reassurance attempt tends to be a token gesture, it is usually insufficient for one to clearly dif-ferentiate a benign state from a strategic malign state at this juncture. Therefore, one should not hastily conclude that the other side is a malign state because it did not respond to one’s initial reassurance gesture.54

At this juncture, one faces a choice: continue or stop. If one is fairly confi-dent about one’s security and really wants to pursue cooperation, one should


go forward with a more credible reassurance gesture (although it should be ensured that such a gesture still will not jeopardize one’s survival if the other side takes advantage of it). At this juncture, it may be desirable to take a dra-matic step to impress the other side and compel it to reveal its true intentions.

After several rounds of reassurance, one should become fairly confident about the intentions of the other state. A genuine benign state that was extremely fearful will have been gradually reassured of one’s benign inten-tions and will now behave like a secure benign state (i.e., reciprocating one’s gestures). In contrast, a genuine malign state will most likely have defected by now, or one would have figured out its true intentions because it has con-sistently refused to reciprocate one’s reassurance gestures. Needless to say, how far one will go with one’s reassurance program will depend on one’s updated assessment of the other side’s intentions after each round of reas-surance. The sooner and more certain one becomes that the other side is a malignant state, the earlier one will halt the reassurance program.

Timing is also important for initiating reassurance. The best time to signal one’s benign intentions toward another state is when it is relatively vulnerable. Hence, when another state is vulnerable, one should not only refrain from taking advantage of its vulnerability but also extend a helping hand. As the conventional wisdom goes, a friend in rainy days is always more trustworthy than a friend in good times, and certainly more trust-worthy than opportunistic predators.

In terms of reciprocating another state’s assurance attempt, two prin-ciples should be followed. First, when a known dove from the other side makes a conciliatory move, one should reciprocate such a cooperative ges-ture. Doing so increases the chance that he/she will stay in office and con-tinue to push for reassurance and cooperation.55 Second, when a known hawk from the other side becomes serious about moving the relationship toward a more cooperative mode, one should reciprocate: When a Nixon flirts, one should play along.

Finally, because uncertainty about others’ present and future intentions is usually permanent, the need for reassurance is also permanent. As a result, one has to constantly deploy reassurance measures to assure others of one’s benign intentions and update one’s assessment about others’ inten-tions. Overall, it takes persistence to succeed in building and sustaining cooperation through reassurance. The way to war may be easy; the way to cooperation is usually hard.

Conclusion

I have tried to provide a more coherent treatment on reassurance as a means for coping with the uncertainty about others’ intentions and building


cooperation. Echoing many others’ views (e.g., Osgood 1962; Glaser 1992; 1994–1995; Kydd 2005), I show that a state can employ a wide range of measures to signal its own intentions and read others’ intentions if the state is benign and determined enough to know others’ intentions and forge a more cooperative relationship with others. Our discussion has important implications for both practice and theory.

Practice wise, our discussion helps us understand many reassurance attempts that are more conventionally known as confidence-building measures (CBMs). Although some CBMs are reassurance attempts with nonmilitary deeds,56 many pundits tend to judge the value of nonmilitary CBMs by their impact on military issues (e.g., arms control). This prac-tice inevitably underappreciates the value of nonmilitary CBMs. While successful nonmilitary CBMs may lead to progress in arms control and reduction in the long run, they may not do so in the short run, because most of them are not designed to have any immediate impact on military affairs. Moreover, the singular focus on the impact of nonmilitary CBMs upon arms control or other military issues by practitioners of CBMs might even have prevented some CBMs from reaping real success. To judge the value of these nonmilitary CBMs, we need to ask whether they have contributed to states’ confidence in each other’s benign intentions, rather than whether they have produced concrete results in arms control and reduction.

By providing a more systematic statement of reassurance as a critical means for building cooperation, our discussion provides states with prac-tical lessons for cooperation-building in the real world. Although states-men have employed many of the measures discussed above to signal their own intentions and read others’ intentions and thus succeeded in building cooperation in the real world, they do not have a thorough understanding of cooperation-building via reassurance. As such, it is perhaps a certainty that states in modern history had many “missed opportunities” to coop-erate with each other (Larson 1997). To prevent more “missed opportuni-ties” for cooperation, reassurance should feature high in the to-do list of statesmen, and our discussion provides a practical guide for them.

Our discussion points to two important directions for future research. The first is the domestic politics of cooperation-building via reassurance. In sharp contrast to the growing neoclassical realism literature on the domestic politics of conflict (Rose 1998; Schweller 2003), the literature on the domestic politics of cooperation has been truly dismal, reflecting both a general bias for war studies in IR and a general structural bias in the study of cooperation-building that was evident in the first generation of the theory of cooperation-building (for a review, see Milner 1992). Although there seems to be a growing interest in the domestic politics


of cooperation recently (e.g., Solingen 1998; Fravel 2005; Schultz 2005), much more need to be done (Tang 2009c).

A second direction is the psychological barriers to reassurance. After Jervis’s two seminal contributions (1970; 1976), students of IR have continued the tradition of borrowing from psychology, especially when it comes to studying conflict (e.g., Stein 1988; Mercer 1996; Crawford 2000; Taliaferro 2004; Tang 2005). In contrast, although social psy-chology has also produced a wealth of understanding on fear, trust, mis-trust, and attribution since the time of Osgood and Deutsch (e.g., Arrow et al. 1995; Ross and Ward 1995; Yamagishi et al 2005; Cook et al 2005), students of IR have yet to pay adequate attention to the important insights from the psychology that may have urgent implications for cooperation-building. It is high time for students of IR to rekindle their interactions with psychology.

Undoubtedly, further research in these two areas should yield impor-tant insights and practical lessons for guiding the crafting and implemen-tation of reassurance.

Part III

Implications and Conclusions


Chapter Six

Recategorizing Realism Theories

Taliaferro (2001) did realism a great service by advancing a fairly sound framework for categorizing realism theories (see also Liu and Zhang 2006). However, because his understanding of the levels of analysis was incom-plete and imprecise and, more importantly, because his understanding of the differences between offensive realism and defensive realism suffers from several critical misunderstandings (chapters 1, 2), his categorization is sometimes misleading if not outright wrong.1 Building upon the discus-sions in previous chapters, this chapter advances a sounder categorization of realism theories.2 Since it will be impossible to give all realism theo-ries their due, at the risk of oversimplification and ruffling some feath-ers, I shall focus on theories that have not been rigorously or correctly categorized.

Here, it is important to emphasize that we have to clearly differenti-ate an author as either a proponent of offensive realism or a proponent of defensive realism from the theories he or she has developed. For instance, whereas Jack Snyder develops an offensive realism theory (a neoclassical offensive realism theory of imperialism expansion), he is a proponent of defensive realism because he draws defensive lessons from the observation that many imperial expansions had ended in tatters.

Theory of Cooperation

According to offensive realism, there is very little common interest among states because states are inherently aggressive and consistently threaten each other. As such, offensive realism in principle rejects cooperation as a viable means of self-help. This essentially means that any theory of cooper-ation is a non-offensive realism theory (Tang 2008a). Theories of coopera-tion can be divided into two large categories: theories based on calculation of mostly, or purely, material interests and theories based on calculation of some kind of ideational interests.


Instrumentalist Theories of Cooperation

Theories of cooperation based on purely instrumental or strategic calculation are defensive realism theories, or “soft realism” theories, as Larson (1997, 3) would like to call them. At its core, these theories espouse essentially an instrumental (and often materialist) approach, and they can be further divided into two strains.

The first strain sees essentially no role or no essential role for institutions, either for the initial forging of cooperation or for the sustaining of coop-eration later on, simply because institutions are unnecessary. Prominent examples of this strain of cooperation theory include Axelrod (1984), Jervis (1978), and Glaser (1994–1995), with the informal approach, and Powell (1991), Fearon (1998), Kydd (2000a; 2005), and Schultz (2005), with the formal (modeling) approach.3

The second strain of theories of cooperation centered on instrumen-tal calculation does see some, and often important, roles for institutions in international politics.4 In these theories, institutions are the means or tools of statecraft toward states’ gains. Accordingly, states decide whether to build institutions almost exclusively on a cost-benefit calculation of (material) interest. States do build institutions when institutions are con-ducive to realizing their goals, but do not do so when institutions are not conducive to, or are unnecessary for, realizing their goals.

Over the years, it has been argued that institutions facilitate cooper-ation by performing many (and often overlapping) functions: providing information, reducing transaction cost, stabilizing expectations, facil-itating commitment, detecting and collectively punishing (thus also deterring) defections (Axelrod and Keohane 1985, 248–254; Keohane 1984, 107; Keohane 1989, 2). More recently, this strain of cooperation theory in which institutions perform important, if not indispensable, role has become more sophisticated. For instance, Ikenberry (2001) argues that international institutions are better means of exercising American power. By channeling power through institutions, the United States exercises stra-tegic restraint. As a result, the United States has been able to make its over-whelming power less threatening and more assuring to other states in the system. Consequently, other states are more willing to work with the United States, rather than try to balance against its overwhelming power. In this context, international institutions are simply a better means for sustaining American hegemony and making the United States a benevolent hegemon (Schweller 2001).5 Gruber (2002) argues that latecomers to institutions flock to institutions to avoid being further marginalized. Weinberger (2003) and Hoddie and Hartzell (2005) emphasize that institutions are a means for signaling benign intentions, reading others’ intentions, and

recategorizing realism theories / 167

institutionalizing cooperation. More recently, Kai He (2008) argues that Southeast Asian states have actually deployed international institutions and organizations as a tool for traditional (soft) balancing, thus “institutional balancing.”

Despite their sophistication, however, these authors’ central thesis that international institutions are a better means of statecraft does not funda-mentally depart from defensive realism. Gruber’s thesis apparently does not differ much from Grieco’s earlier thesis that the Maastricht Treaty was a binding mechanism for smaller European states to improve their positions relative to the bigger players (e.g., Germany, France, and Britain) through the option of “voice” (Grieco 1996). Likewise, Weinberger’s thesis and Hoddie and Hartzell’s thesis are no different from Kydd’s costly signaling theory (Kydd 2005). Similarly, Ikenberry’s call for self-restraint is no dif-ferent from discussions by Kydd (1997b, esp. 139–147) and Tang (2004, esp. 6, 27–28), as well as earlier discussions by Wolfers (1952, 494–497), Jervis (1978), and Glaser (1994–1995, esp. 67–70). Finally, Kai He explic-itly admits that the logic of “institutional balancing” falls squarely within the logic of (defensive) realism.

In the heated debate between neorealism (which generally does not see any significant role for institutions) and neoliberalism, many have failed to recognize that just because institutions do play some role in interna-tional politics it does not automatically mean that realism is discredited. Many realists do see a role for institutions, if these institutions serve states’ interests, especially material interests. In fact, when institutions perform the above-mentioned functions, institutions do not contradict defensive real-ism’s understanding of cooperation at all, because institutions are merely a means toward the ends. These “institutionalism” theories are entirely con-sistent with realism’s core assumption that states are strategic actors, and these theories fall squarely within a (defensive) realism approach (Krasner 1982a; Jervis 1999).

Theory of Cooperation Based on Ideational Interest

Realism asserts that states will obey some rules and norms only when they believe that obeying such rules and norms brings tangible (material and psychological) payoffs.6 As such, only theories of cooperation that explic-itly argue that states obey certain institutions (or rules) and norms because they take these rules and norms themselves to be interest can qualify as non-realism theories of cooperation.

Because it is difficult to find any state (or any human) behavior that is not influenced by some form of calculation in material interest, the bar for a theory to quality as a nonrealist theory of cooperation should not be set


too high. Perhaps as long as a theory of cooperation conforms to the min-imal conditions below, it qualifies as a non-realism (i.e., either institution-alism or constructivism) theory of cooperation. In other words, material power cannot be the whole calculation, nor should it even dominate states’ calculations. Rules or norms as codified ideas must have an impact upon state behaviors, independent of material calculations.

Obviously, to support such a claim, the most convincing evidence is that states have refrained from overthrowing certain rules or norms even if they had the power to do so, and even when doing so would have brought them tangible material benefits. In other words, one must show that rules and norms have a restraining effect on power and material calculation, that is, states are willing to defend and obey certain rules and norms even if doing so will jeopardize some of its material well-beings and they have the power to disregard those rules and norms (Krasner 1982a; idem, 1982b; Jervis 1999).7

Understood as such, when it comes to international politics, the differences between institutionalism and constructivism on one side and (defensive) real-ism on the other are far more profound than Keohane has asserted. At the very least, it is misleading for Keohane to argue that the debate between the rationalist approach and the constructivist approach constitutes the major debate in theories of international politics (Keohane 1989, Chapter 7).

This is so because even within a rationalist approach (i.e., where states are strategic actors), one can still subscribe to either materialism or ideationalism.8 Subscribing to materialism, realism insists that ultimately it is material calculation that determines states action. In contrast, sub-scribing to ideationalism, institutionalism insists that material calculation does not necessarily determine state behavior, and that institutions, as embodied ideas, can also determine state behavior. Constructivism goes further, insisting that even ideas that are not institutionalized (i.e., identi-ties) determine state behavior.

After delineating institutionalism and constructivism into the ideation-alism camp, we can better understand the differences between realism on the one side and institutionalism and constructivism on the other. Institutionalism accepts constructivism’s first claim that ideas determine behaviors but reject its second claim that states’ behaviors are constructed (or sociological and reflective, according to Keohane 1989) rather than strategic. Within institutionalism, states remain strategic actors, but only because they believe that rules and norms are a type of interest. In con-trast, realism rejects not only constructivism’s first claim that ideas deter-mine states’ behavior, but also constructivism’s second claim that states’ behaviors are constructed rather than strategic (Wendt 1995, 81).

To summarize, then, realism has problems with constructivism not on just one but two fronts: material versus ideational, rationalist versus


constitutive. In contrast, institutionalism has problems with construc-tivism only on the rationalist versus constructed front. Contra Keohane, the debate between realism and neoliberalism/constructivism is thus more than just a debate between the rationalist approach and the constructivist approach. Proclaiming that the debate between the rationalist approach and the constructivist approach constitutes the main battle in IR theory denies a place for realism and materialism in IR theory.9

Other Post-Waltz Realism Theories: A Brief Sorting

The Security Dilemma: Doubters versus Usurpers

As shown in chapters 2 and 3, the existence of the security dilemma is indeed conditional rather than universal. As such, contra Taliaferro (2000–2001), the two previous challenges to the thesis that the security dilemma is universal and it has been a major cause of war are actually defensive realism theories (Schweller 1996; Kydd 1997b).10

Meanwhile, although both Copeland (2000) and Mearsheimer (2001) have attempted to co-opt the security dilemma, this co-opting of the security dilemma does not change the offensive realism nature of their theories (Tang 2008a).

Theories of Alliance

Realism, whether offensive or defensive, expects states to balance against threats (Waltz 1987; Mearsheimer 1995, 12; 2001, 33). As such, contra Taliaferro (2001, 135, table 6.1), theories of alliance when facing a com-mon and clear threat, prominently represented by Walt (1987) and Glenn Snyder (1997), are not defensive realism theories per se. These theories are consistent with both offensive realism and defensive realism. This is so even though proponents of defensive alliance such as Walt and Glenn Snyder are defensive realists because they are generally critical of offensive realism and have consistently advocated defensive realism policies of self-restraint (e.g., Snyder 2002; Walt 2005).

Similarly, the implicit message for alliance in Waltz’s theory of balanc-ing of power is also consistent with both offensive realism and defensive realism, not in the least because Waltz’s balancing against power does not take states’ intention into account (Waltz 1979).11

Theories of Power Transition and (Preventive) War

Correcting Brooks’s error of classifying Gilpin’s theory as a defensive realism theory (Brooks 1997), Taliaferro (2001, 135) correctly classifies


Gilpin’s theory of hegemonic theory of war and Organski and Kugler’s theory of power transition and war as offensive realism theories (Gilpin 1981; Organski and Kugler 1980).12 Indeed, the two theories are essen-tially identical. Both argue that a rising power almost inevitably challenges the reigning hegemon, thus leading to a hegemonic war. Both theories also imply that peace is more likely under hegemony or unipolarity.

Table 6.1 Post-Waltz realism theories at different levels of analysis. Besides drawing partly from Table I in Taliaferro (2000–2001), I have greatly expanded the content on offensive realism and defensive realism in this table. I have added a column on neoliberalism/institutionalism theories, although many existing institutionalism theories are actually defensive realism theories.

Offensive Realism Defensive Realism Neoliberalism

Theories at the Structural LevelMearsheimer: tragedy of great power competitionGilpin/Organski-Kugler: theory of power transition and hegemonic warCopeland: theory of dynamic differentiation and preventive warWohlforth: theory of unipolar stability and hegemonic policy

Herz/Jervis: the structural origin of the security dilemmaAxelrod, Glaser, Jervis, Lipson, and Powell: cooperation as self-help and conditions for cooperationGlaser, Kydd, Schweller: conditions of the security dilemma and its implicationsBull: theory of anarchical society as order

Keohane and Nye: interdependence and powerKeohane: theory of institutional stability after hegemonyDeutsch/Hass: European integration and peace

Theories at the State Level (neo-classical theories) Zakaria/Snyder: neoclassical theories of imperialism and expansionLabs: theory of expanding war aimsHager and Lake: theory of competitive decolonization

Glaser: theories of rational and irrational arms racesKydd: theories of trust and cooperationIkenberry/Gruber/Weinberger: institutions as strategic restrains and signaling deviceSchultz: theory of risking cooperation and peace

Putnam: theory of two-level game and cooperationSimmons: theory of international commitment and complianceSolingen: theory of economic liberalization, domestic coalition, and nuclear restraint

Theories at the Individual Level: human nature or nonhuman natureTaliaferro: prestige, reputation, loss aversion, and war in the peripheryMarkey: prestige and warThayer: conflict for fitness

Young: intellectual and entrepreneurial leadership in institutional building.


Taliaferro (2000–2001), however, mistakenly classifies Copeland’s dynamic differentials and preventive war as a defensive realism the-ory. Taliaferro’s mistake is perhaps due to the fact that Copeland (2000, Chapter 1) explicitly integrates the security dilemma into his theory and Taliaferro believes that whether a theory admits the security dilemma is a real fault line between offensive realism and defensive realism.

Looking closely, Copeland’s theory must be classified as an offensive realism theory, for three major reasons. First, Copeland’s theory explicitly synthesizes the theory of power transition and the theory of preventive war into a more dynamic theory of great-power war. Thus, Copeland’s theory is essentially a theory of preventive war.13 Yet, as shown in chap-ter 4, preventive war is generally inconsistent with the logic of defensive realism. Second, it is apparent that Copeland actually has a spiral rather than a genuine security dilemma in mind when he talks about the secu-rity dilemma, because the security dilemma for Copeland merely requires security-seeking motive to operate (see chapter 2). Among the dyads Copeland examined, one or both sides were intentionally aggressive, and there was no security dilemma between them. Third, in Copeland’s lad-der of strategies (see chapter 4), cooperation is nowhere to be found, and reassurance is defined as appeasement (worse than do-nothing!). As a result, Copeland’s theory has a strong competition bias that is the hall-mark of offensive realism. Most importantly, intentions have no role in states’ calculation of preventive war in Copeland’s framework. As such, Copeland also implicitly assumed that states assume the worst over oth-ers’ intentions, and this is a fundamental assumption of offensive realism (Tang 2008a, 466).

Theories of Expansion and Imperialism

Taliaferro (2001, 135) correctly classifies Zakaria’s state-centric realism the-ory of expansion as an offensive realism theory (Zakaria 1998). Taliaferro, however, fails to classify Jack Snyder’s logrolling theory of imperialism as a quintessential offensive realism theory (Snyder 1991). Yet, Snyder’s the-ory essentially seeks to explain imperialism with domestic politics, just as Zakaria’s theory does. As such, Snyder’s neoclassical theory of imperialism is an offensive realism theory, although he draws defensive realism lessons for states.14

Theories of Domestic Mobilization and Constraint

When facing a clear and present (perceived, rightly or wrongly) threat, both offensive realism and defensive realism predict (and recommend)


that a state will try to balance against the threat vigorously. Because a major means of internal balancing involves domestic mobilization, the-ories of domestic mobilization are consistent with both offensive realism and defensive realism, if the theories deal with a situation where a state mobilizes its resources to face a clear and present danger.

Christensen (1996) examined how Truman exploited international opportunities (i.e., crises) to overcome domestic constraints and mobilize national resources for a more hardened containment approach against the former Soviet Union.15 Looking at domestic constraint, Schweller (2006) examined why Britain and France failed to adequately balance the looming threat from Hitler’s Germany before WWII. Neither Christensen’s theory nor Schweller’s theory can be neatly categorized into either offensive real-ism or defensive realism (see also Rose 1998, 161–164).

Theories of Costly Signaling

The literature on costly signaling can be divided into two major strains. One strain of the literature focuses on signaling one’s own benign inten-tions and knowing others’ intentions and then building trust to move toward more extensive cooperation when possible. This group of litera-ture considers building peace through cooperation as a problem of sig-naling benign intentions and building trust. Because this strain of the cost-signaling theory focuses on cooperation, it falls within the defensive realism camp. Another strain focuses on signaling resolve in conflict (war or threat of war), essentially taking conflict as a bargaining situation and focusing on how to win the conflict by signaling resolve. This literature presumes that two parties are already in an irreconcilable conflict and the task is to win the conflict, not to resolve it through cooperation. Because this strain essentially takes conflictual situations as given, it is consistent with both offensive realism and defensive realism.16 A prominent exam-ple of this strain is the rational deterrence theory (for a good review, see Jervis 1989).

Understanding Waltz

In his classic work, Waltz (1979, 105–106; 1986, 336) has generally ruled out cooperation as a legitimate means of external self-help. This would have put him in the offensive realism camp. Waltz (1979, 91, 118, 126–127), however, also explicitly emphasizes that maximizing relative power may or may not serve security. Such a stand would then put Waltz in the defensive realism camp because almost all defensive realists contend that maximiz-ing relative power may or may not serve security (e.g., Glaser 1994–1995,


71–72; Haas 1953, 466; Lynn-Jones 1998, 177n47; Wolfers 1952, 490, 496–498), whereas offensive realism believes that maximizing relative power is the only viable means of external self-help and is always condu-cive to security (see chapter 1). Hence, Waltz’s theory cannot be neatly categorized into offensive realism or defensive realism, and this might have contributed much confusion.17

Ultimately, though, Waltz should be considered as a defensive realist because he does have a strong defensive, or status quo, bias, as Schweller (1996) puts it (see also Mearsheimer 2006, 109–110). Over and over, he has prescribed policies with self-constraint and moderation, from the Vietnam War, to the nuclear weapon debate, and to the expansion of NATO (Waltz 1967; 1990, 2000).18

Classical Realism and Pre-Waltz Modern Realism

For convenience, I use classical realism to denote all realism theories before the coming of Reinhold Niebuhr and E.H. Carr, and I use pre-Waltz mod-ern realism to denote realism theories after the coming of Niebuhr and Carr but before the coming of the Waltzian structural (behaviorist) revo-lution. Because classical realists and pre-Waltz modern realists wrote their major treatises before the behaviorist revolution in social sciences, most of them did not state their core assumptions explicitly, and their logical elab-orations were often less than clear-cut and consistent. Moreover, realists then were mostly in conflict with idealists and tended to overlook the dif-ferences among fellow realists. This poses a problem for classifying them into offensive realists or defensive realists.19 Nonetheless, with the more rigorous formulation of defensive realism (and offensive realism) above, a reasonable categorization of some classical and pre-Waltz modern realists is possible.

Classical Offensive Realists: Shang Yang, Kautilya, Han Fei Tzu, and Machiavelli

In Western realism, the first quintessential offensive realist was undoubt-edly Machiavelli.20 For him, an ideal prince is someone who can conquer and harvest honor like the Roman Empire.21 In addition, Machiavelli actu-ally dreaded peace and called for a return to a past and bloodier world than the one he lived in, as he complained that Italian soldiers had lost their primitive warrior spirit and that most princes were not as Machiavellian as he had hoped.

Yet, long before Machiavelli, three prominent offensive realists had already achieved fame in other parts of the ancient world. These offensive


realists not only advocated but actually practiced offensive realism and succeeded splendidly! They are Shang Yang (390–339 B.C.) and Han Fei Tzu (280–233 B.C.) of ancient China, and Kautilya (~300 B.C.) of ancient India.

Shang Yang was the advisor to King Hui of the Kingdom of Qin, which eventually unified the heartland of today’s China and formed the first Chinese empire in 221 B.C. For Shang Yang, states are inherently aggressive and their aggressiveness is limited only by their capabilities (Shang Yang ~339 B.C., Vol. 7). As such, states have either to expand or be expended. For Shang Yang, the business of a state is about two things and two things only: “Nong” (agriculture to raise revenue) and “Zhang” (war for conquest), and the former is simply a means to build the mate-rial foundation for achieving the latter (Shang Yang, Vol. 3).22 In such a world, preventive wars are not only just but also should be pursued with earnestness (Vol. 4). Han Fei Tzu, who was an advisor to a latter King in the Kingdom of Qin, essentially adopted and developed Shang Yang’s theses to provide a more comprehensive treatise on politics, both domestic and international.23 His thesis was also undoubtedly offensive realism.

Writing after Shang Yang, but before Han Fei Tzu and long before Machiavelli, Kautilya promulgated roughly the same doctrine of offensive realism in his Arthasastra (Kautilya 1969). Kautilya also saw the world as a world in which one has to either expand or be expended (Boesche 2003, 19). In such a world, every state must expand to maximize (relative) power, and attack whenever the opportunity was propitious. Kautilya thus “preaches an ideal of conquest” (R.P. Kangle, quoted in Boesche 2003, 28), just like Machiavelli preached more than a millennium later.

Social Darwinism, Geopolitics, and Fascism

Social Darwinism and Geopolitics are offensive realism schools.24 Social Darwinism gives geopoliticians the supposedly “scientific” foundation for constructing a scientific theory of power politics based on organisms and “survival of the fittest.” As Karl Haushofer, the founder of Geopolitics, claimed: “Geopolitics is the scientific foundation of the art of political action in the life-and-death struggle of state organisms for Lebensraum (Haushofer, quoted in Schweller 1996, 96).”

Social Darwinism and Geopolitics explicitly called for control through expansion and conquest. While Mahan, Mackinder, and Spykman had debated on which part of the earth was the most critical part to domi-nate (i.e., the sea, the Eurasia heartland, or the rimland), all geopoliti-cians agreed that states seek domination toward regional, if not global,


hegemony, and that (as such) states are in continuous conflicts with each other (e.g., Spykman 1942, 21–25). Obviously, this overall position is little different from one of modern offensive realism’s key dictums (i.e., Mearsheimer 2001).

With crude (and wrong) understanding of social evolution, Social Darwinists preached acquiring “living space” through expansion and conquest, because states have to grow constantly. Thus, a minister to Catherine the Great asserted, “That which stops growing begins to rot” (quoted in Jervis 1978, 186). In England, the Anglo-German competition was depicted as “the first great racial struggle of the future: here are two growing nations pressing against other . . . all over the world, One or the other has to go; one or the other will . . . go.” Likewise, in Germany, Social Darwinists asserted, “If Germany does not rule the world. . . . it will disap-pear from the map; it is a question of either or” (quoted in Van Evera 1984, 66). For Social Darwinism and Geopolitics, expansions are preordained for states.

Some Social Darwinists go further by calling for the extinguishing and enslaving of other peoples based on ethnic superiority. These individu-als were sheer racists and they eventually became Fascists. For instance, Fredrick von Bernhardi (1914) explicitly called Germany to expand and dominate others to have more living space because culturally superior peo-ple should dominate others, and Germans had cultural superiority (as cited in Jack Snyder 2003, 358). Fascism is essentially a racism-charged form of offensive realism (Schweller 2009). Undoubtedly, social Darwinism and Geopolitics provided much of the intellectual foundation for expansion-ist policies among European states, and was a major cause of WWI and WWII (Van Evera 1984, 62–63; Schweller 2009).25

The First Quintessential Defensive Realist: Arnold Wolfers

Arnold Wolfers should be considered the first quintessential defensive real-ist. In his classical article (Wolfers 1952), Wolfers stated three principal points. First and foremost, while it was Butterfield and Herz who first coined the term of “security dilemma,” it was Wolfers who first explicitly articulated the defensive implications of the security dilemma (494–497). Wolfers recognized that self-restraint and taking others’ interest into con-siderations should be the choice if one accepts the existence of the security dilemma. Further, Wolfers emphasized that states must face the trade-off of “the efforts to change the intentions of an opponent” and “the efforts to build up strength against him,” and that trade-off cannot be avoided “if security policy [is] to be expedient” (497; emphasis added).26


Second, Wolfers pointed out that national security is an ambigu-ous symbol because different individuals can define it very differently. Moreover, Wolfers recognized that “very high security aspirations tend to make a nation suspect of hiding more aggressive aims” (488, 492–494).27 Wolfers further noted that idealistic statesmen who go to war for ideo-logical reasons (i.e., spreading liberty) and imperialist countries that are out for conquest and glory are fundamentally similar. They are all “mad Caesars” or offensive realist states (489, 492).

Finally, Wolfers not only highlighted that Morgenthau’s point—that politics among nations is all about power—can only be correct if power and security are “synonymous” (Wolfers 1952, 485n4), but also explic-itly argued that power and security should not be “synonymous” and that power accumulation can be self-defeating due to the existence of the secu-rity dilemma (490, 494; see also chapter 1 this volume).

Concluding Remarks

In this chapter, I build upon Taliaferro and attempt a more rigorous cate-gorization of realism theories. From the preceding discussion, it becomes obvious that some of our understanding about realism theories has been incorrect. Three major causes lie behind our misunderstanding.

First, many realism theories have not been formulated explicitly enough because they often require hidden assumptions to operate. Moreover, some seemingly coherent bodies of theories actually consist of several components that are inconsistent with each other (a prominent example is ODT).

Second, many have conflated the nature of a theory with the policy implications of the theory and/or the policies advocated by the author of the theory. Because most realists tend to separate their moral concern and policy commendations from their theoretical constructs (Desch 2003, 419–421), scholars’ specific policy recommendations may or may not be derived from their scholarly writings. As such, just because a scholar gen-erally proposes a defensive realism policy, it does not necessarily mean that his theory is a defensive realism theory. Similarly, just because a scholar develops a defensive realism theory does not necessarily mean that he or she is a defensive realist.28

Finally, some theories cannot be conveniently classified into either offensive realism or defensive realism theories simply because these theo-ries are consistent with the core assumptions of realism. Such theories are simply realism theories.

Overall, classifying theories, but especially classifying a scholar as an offensive realist, a defensive realist, or a neoliberalist according to his or


her theoretical writings, must be done very cautiously. Some scholars (e.g., Waltz) have been inconsistent with their arguments and some have con-tributed ideas to more than one grand theory (e.g., Schweller), and it is difficult to classify them unless they identify themselves with one school or another. Moreover, even scholars themselves may classify themselves incorrectly (i.e., Schweller on his work on the security dilemma).


In Lieu of Conclusion: Policy Implications

As students of realism, we should resist the temptation to advocate for policy recommendations without discretion. Yet, we should prescribe policy recommendations when the theoretical framework we have devel-oped bears clear implications for the real world. After all, realism has always been both a descriptive and a prescriptive paradigm. After advancing a more coherent statement of defensive realism in the preceding chapters, I shall now venture to prescribe some broad guidelines for seeking security in the world today, because defensive realism does hold important practical lessons for managing security in the world today.

Before I proceed, however, three caveats are in order. First, although the policy prescriptions below are explicitly geared toward improving states’ welfare via minimizing conflict and securing peace, the prescriptions are based on a social evolutionary reading of the history of international politics (Tang 2009b) and a normative preference for peace, rather than a belief that defensive realism is timeless theory that can cover the whole history of international politics.1 Second, contra some (realists) who tend to prescribe policies that may not be totally consistent with their theoret-ical arguments, I shall only prescribe policy prescriptions that are consis-tent with the theoretical expositions developed in the preceding chapters. Hence, while I certainly believe that certain aspects of international pol-itics can be better managed with an approach that goes beyond defensive realism, I shall limit my prescriptions to those that are firmly within the territory of defensive realism. Third, since I have outlined a more specific operational code for a defensive realist state in chapter 4, I shall refrain from drawing more specific policy implications here (e.g., when facing an offensive realist state). Doing so will simply be a repetition of what I have said in the preceding chapters.

Security-seeking in a Defensive Realism World

The world has firmly evolved from an offensive realism world into a defensive realism world (Tang 2009b). Most states—including most


regional great powers (but see the next section)—in today’s world have been selected and socialized into defensive realist states. After WWII, outright territory expansion has been rare and is essentially considered illegitimate. As a result, the rate of state death has been remarkably low, and violent state death has virtually ceased after WWII (Zacher 2000; Fazal 2004; see also Waltz 1979, 137). Moreover, the world is unlikely to revert to an offensive realism world that had existed in the past. As a result, our world really is and will remain less dangerous and more peaceful than it used to be. In a defensive realism world, defensive realism should be a fitting theory for guiding states’ security policies today and in the near future, although the exact length of this time frame cannot be precisely specified because we do not know when a defensive realism world will be transformed into an even more peaceful neoliberalism world.

Moreover, this does not mean “the end of [international] history” as we know it. As long as states live under anarchy in which resources remain limited, competition among states will remain a fact of life and real conflicts of interest among states will emerge here and there. Moreover, offensive realist states—although rare—may still emerge (e.g., Iraq under Hussein, the United States under George Bush), and the fear that offensive realist states may still emerge means that most states will have to maintain a military establishment that is capable of defending themselves (and thus harming others at the same time). As a result, actual conflicts or dangerous confrontations may still emerge. If this is the case, states will still have to retain a healthy dose of defensive realism for managing their security for a while.

From the preceding chapters, we can derive two general principles for seeking security in a defensive realism world. First and foremost, states shall not automatically assume the worst over others intentions, although the possibility that the other side can be an offensive realist state should not be neglected. The worst thing that states can do in a defensive realism world is to assume that other states are inherently aggressive, because doing so inevitably exacerbates fear and thus the security dilemma. Instead, states can afford to fear each other a bit less—perhaps much less—than they do in an offensive realism world.2

Second, when many, if not most, states in the system are defensive realist states, cooperation among states is not only desirable but also highly possible. Through reassurance, states can signal their own benign inten-tions and read others’ intentions. Along the way, they can develop piece-meal cooperation to build more trust and gradually move the relationship toward a more cooperative status. And whenever possible, states should institutionalize their cooperation as much as they can and use organiza-tions to enforce them.3 As affirmed in chapter 6, institutionalizing existing

in lieu of conclusion / 181

cooperation so that such cooperation can last longer and future coopera-tion becomes easier is inherently consistent with defensive realism.

If states follow these two broad prescriptions, the defensive realism world in which we live today may become even less dangerous than it has been.

Living with a Sometimes Offensive Realist Superpower

Although our world is a defensive realism world, it seems that the United States—the most powerful state in today’s international system—has largely remained, or has become, an offensive realist state after the cold war, and especially after September 11 under George W. Bush.4 The behaviors of the United States after the cold war have certainly been more consistent with offensive realism on several fronts. Foremost, the Defense Policy Guidance (DPG) 1992–1994 reads like a doctrine that is straight out of an offensive realism textbook. In it, Dick Cheney, Paul Wolfowitz, and their associates stated explicitly that the goal of the United States is to preserve its primacy by resorting to preventive measures, if not preventive war, to prevent other countries from rising, thus prolonging the unipolar moment.5 Of course, the doctrine of preventive war is the hallmark of an offensive realism doctrine (Mearsheimer 2001; Tang 2008a).

Second, the military behaviors of the United States after the cold war have largely followed the Cheney-Wolfowitz DPG mentioned above, despite different administrations taking power. The United States has not significantly reduced, but has actually increased, its defense spending relative to other states.6 It has also expanded rather than shrank its sphere of influence through the expansion of NATO and U.S. military presence across the globe (Johnson 2004).

Third, and perhaps most alarming, after September 11, the United States had not only explicitly outlined a doctrine of preventive war in the Bush Doctrine (which intentionally misrepresented “preventive war” with “preemptive war”), but had also actually launched such a war against Iraq. The “Bush Doctrine,” by explicitly endorsing preventive war, is indisputably following a doctrine of offensive realism.

Finally, the United States has also taken on the task of spreading democracy by force. Of course, when a state is out to spread its ideology by force, it is no different from “mad Caesars” who are out for conquest and glory at any price. They are all imperialists or offensive realist states (Wolfers 1952, 492).

Overall, there are ample evidences that the United States was an imperialist, a revolutionary, or an offensive realist state, especially under George Bush and his neo-con supporters (Ikenberry 2002; Jervis


2003; 2006; Hendrickson 2002; Cox 2004; Layne 2006b; Nelson 2005; Offner 2005; Prestowitz 2003).7 Perhaps even more dishearteningly, largely due to its unique geographical location and overwhelming relative power over other states, the United States may well remain an offensive realist state for the foreseeable future, although it may recoil a bit after a major military debacle (say, in Iraq).

Because the United States will remain the most important player (i.e., a hegemon or semi-hegemon) in the international system for the foresee-able future, the fact that it may well remain an offensive realist state poses an enormous challenge for other (defensive realist states) in their seeking security within the system.

Under the unipolar world, directly confronting the lone superpower with military means is not a viable option. Major regional powers (and other states in the international system), however, can perhaps try to form “an axis of constraining” to constrain the lone superpower. Because the United States will almost inevitably demand some support from these other states, they can collectively tell the United States that some of its actions are not only damaging others’ interests, but most importantly are also hurting its own interests.8

Because other states cannot balance against the United States militarily, however, ultimately we can only hope that the United States will eventu-ally learn the lesson that being an offensive realist state no longer pays and becoming a defensive realist state is the way to go. And when the United States does become a defensive realist state, the cardinal principle for guid-ing U.S. security strategies may be to behave with self-restraint, when the external restraint is no longer as robust as it was under bipolarity, as many have argued (e.g. Nye 2001, xii, 16–17, 25–26; Taliaferro 2000–2001, 159–160; Jervis 2006; Walt 2005).

For this, American scholars and opinion makers within the defensive realism and other non-offensive realism camps have an “enviable” task in front of them. They not only need to expose the fallacies of offensive realist policies in a defensive realism world, but also need to reveal the benefits of defensive realist policies in a defensive realism world. Most importantly, they need to engage the American public and elite in a more forceful and sustained way to change course. Defensive realists should welcome such a task. After all, realism is as much about theorizing international politics as about practicing international politics realistically.

Conclusion, Sort of

So, what does this world lead us? We are living in a defensive realism world in which cooperation can be quite possible and fruitful. Unfortunately,

in lieu of conclusion / 183

this world will not become a paradise any time soon, if ever, due to states’ fear of each other and the presence of a sometimes offensive realist super-power. Instead, states within this international system have to live under a constant though limited uncertainty over each other’s intentions and the fear derived from this uncertainty.

This uncertainty over each other’s intentions and the fear derived from this uncertainty will not go away, and they make living in our world interesting but potentially dangerous. Fortunately, the future lies within our, rather than in aliens’, hands.


Appendix I: World War I and the Cold War Revisited

World War I (WWI) and the cold war remain two of the most important cases for students of international politics. This appendix provides a more definitive interpretation about the nature of the two conflicts, in light of the more rigorous understanding about the security dilemma and the spiral developed in chapter 2.

WWI as a War of Spiral, not a War of Security Dilemma

Almost a century after WWI, scholars are still debating the nature of the causes of this war. One school argues that WWI was “an inadvertent war” primarily caused by a security dilemma or a spiral,1 whereas another school argues that WWI was “an inevitable war” that was mostly caused by the belligerence among the states, especially Germany (for a recent exchange, see Lieber 2008; Snyder and Lieber 2008). The rigorous differentiation of the spiral model from the security dilemma model points to a more defin-itive interpretation of WWI: WWI was not a war that was mostly driven by a security dilemma, period.

To begin with, social Darwinism and Geopolitics, the prevailing ideol-ogy of that time, preached an overtly aggressive approach toward security, and this ideology was believed by most of the major states in the War, including Germany, France, Britain, Russia, and Japan (Van Evera 1984, 62–63, 66–67). As noted in chapter 6, this ideology belongs to offensive realism in the extreme.

Second (and not surprisingly), major states of the war were all impe-rialist powers or offensive realist states. These states were intentionally threatening each other, and there was no genuine security dilemma among them.

Here, it must be emphasized that while existing literature has empha-sized the aggressive intentions of Germany and Austria-Hungary, it has underemphasized the aggressive intentions of the other major participants in the war. Yet, as Schroeder (2005) underscored, almost every major state

186 / appendix i

in the war was expanding somewhere, and their competition for power, prestige, and colonies had already put them in a collision course with each other. Britain was expanding in East Africa, South Africa, Southeast Asia, and Central Asia. France was expanding in West Africa, Central Africa, North Africa, and Southeast Asia. France was also eager to take revenge on Germany and regain Alasace-Lorraine. Austria wanted to maintain its fragile empire through expansion in the Balkans. Russia was expanding in East Asia and Central Asia. Finally, Japan entered the War mostly because it wanted to expand its conquest in East Asia. The United States might have been the only major state in the war that had no direct territorial con-flict with other major states, but only because it was expanding in Central America, the Caribbean, and the Western Pacific, areas where other great powers were not engaged.

Overall, WWI was fought between two offensive alliances that were composed of offensive realist states, and the dynamics of security dilemma could not possibly have played a major role in causing the war (Snyder 1985, 176–178; see also chapter 2 this volume). Considering the objective and subjective irreconcilable conflict of interest among the major states, the war was inevitable, literally. While some inadvertent events (e.g., the assassination of the Archduke, Bethmann’s duplicity in relaying Kaiser’s message to Vienna on July 28, 1914) had contributed to the making of the war, they merely influenced the timing of the war.

Of course, the fact that WWI was not an inadvertent war does not mean that spiral dynamics were not at play. Major states’ belief in Social Darwinism/Geopolitics and the fear of relative decline on the part of Germany and Austria-Hungary had played important roles in driving up the spiral of 1914.

The Cold War: The Zigzag of Spiral and Security Dilemma

As World War II was approaching its end, many American decision- makers retained a generally benign image about Stalin’s Soviet Union and genuinely hoped for a postwar international system that would be managed by the “Four Policemen” (the two superpowers plus Britain and China). Many American decision-makers also worried about the danger of an avoidable confrontation between the two superpowers driven by a security dilemma (Kydd 2005, Chapter 4; Weinberger 2003).

Because Stalin’s Soviet Union was a genuine expansionist, however, there was only a spiral of the expansionist kind of threat, but no real security dilemma between the two superpowers. The United States was imaging a security dilemma.

appendix i / 187

Of course, it did not take long for the United States to realize the reality. Through a series of reassurance attempts that were explicitly designed to reassure the Soviet Union and at the same time to gauge its intentions, and aided by Stalin’s expansionist behaviors, the United States quickly came to view Stalin’s Soviet Union as aggressive (Weinberger 2003; Kydd 2005, Chapter 4). The Iranian Crisis in early 1946 was the final nail on the coffin, and the Korean War completely eliminated the two sides’ uncertainty about each other’s intentions.

After Stalin’s death in 1953, both Beria and Malenkov advocated for “peaceful co-existence.” More importantly, after consolidating his grip on supreme power, Khrushchev explicitly repudiated some of Stalin’s expansionist policies and strived to implement “peaceful co-existence” with the United States and its allies. Here, the cold war was transformed from a spiral into a security dilemma.2 Yet, because it emerged from an expansionist threat, the security dilemma inherited by Khrushchev and Eisenhower (and later, Kennedy) was a very “deep” one.

After Khrushchev was removed from power, Brezhnev took power and promulgated the “Brezhnev Doctrine,” and the Soviet Union reverted to being an offensive realist state. As a result, the cold war reverted to an expansionist threat. In 1982, Ronald Regan came to power and the United States also became an offensive realist state. The cold war then became a “mutual threat” or “deadlock.”

After Gorbachev came to power and launched glasnost and perestroika, the Soviet Union became a defensive realist state again. At this juncture, the cold war again reverted to an expansionist threat, the only change being that the more aggressive side was Regan’s America this time. When the more moderate George H. Bush took power in 1989, the cold war was transformed back to a genuine security dilemma. Because of Gorbachev’s extraordinary effort, the cold war finally came to an end.3

Overall, it is simplistic to ask whether the cold war was indeed a security dilemma. Posing the question as such essentially makes the question intractable. A more careful reading of the nature of the cold war points to the fact that the nature of the cold war underwent several changes.


Appendix II: Liber als, Conservatives, Doves, and Hawks

The rigorous categorization of realism theories (and realists, to a less extent) provided in chapter 6 also allows us to better understand the often confusing labels of liberals, conservatives, liberal doves, liberal hawks, “paleo-cons,” “realcons,” and “neo-cons” in the current American political scene (Desch 2001).

A prevailing myth in the American political discourse is that liberals and hawks are simply incompatible: liberals are generally doves, and only conservatives can be hawks. The emergence of “neo-cons” or “liberal hawks” after the terrorism attack of September 11 finally made the myth no longer tenable.

In principle, belief in the moral righteousness and inherent peacefulness of liberal democracy is a belief in a particular domestic political system, and it does not mandate a particular belief for conducting international poli-tics. As such, individuals who believe in the virtue of liberal democracy can be offensive realists, defensive realists, neoliberalists, and constructivists when it comes to international politics.

And this is indeed the case. Some believers in liberal democracy respect other states’ sovereignty in choosing their regime types and are against exporting any ideology by force. These individuals are more akin to defen-sive realists. In contrast, some believers in liberal democracy do believe in and may even preach the spreading of democracy through force. These liberals are offensive realists or “mad Caesars” (Wolfers 1952, 492). For realism, exporting democracy using violent means is no different from exporting dictatorship by violent means.

Thus, among liberals, only those who do not believe in exporting (or imposing) liberal democracy by force are defensive realists when it comes to international politics. In contrast, liberals who do believe in export-ing (or imposing) liberal democracy by force—however reluctantly—may be liberals, but they are also offensive realists. Examples of these liberal hawks are Michael Ignatieff and Anne-Marie Slaughter, both were strong

190 / appendix ii

proponents of preventive war and staunch supporters of the recent U.S. invasion of Iraq.

The fact that some liberals are also offensive realists should not come as a surprise. After all, there has always been an imperialistic or illiberal legacy within international liberalism (Jahn 2005; Waltz 2000, 11–18). When Britain was the preeminent imperialist power, many British liberals, from John Stuart Mill, to Lord Palmerston, to Lord Salisbury, to Winston Churchill, were all strong advocators of imperialist expansion and gunboat diplomacy. Not surprisingly, these liberal imperialists often covered their imperialistic agenda with highly moralistic and evangelistic tones such as liberation, civilizing, and spreading Christianity (i.e., “White Men’s Burden”). Their belief in the moral superiority of democracy provides the moral justification for their imperialist agenda. Back then, these liberal imperialists were called the “Limps.”1 Today, with the United States being the dominant state, the home base of liberal imperialism has shifted to the United States. The “neo-cons” (or neo-imperialists) in the United States are only “neo” in the sense that they now seek cover under the banner of spreading democracy and freedom (rather than racial superiority and spreading civilization). Yet, the thrust of the neo-imperialists’ argument is not much different from that of the old liberal imperialists.2

Understood correctly, the difference between “liberal hawks” (i.e., “neo-cons”) and traditional hawks (i.e., “paleo-cons” or “realcons”) is two-fold. First, “neo-cons” believe in transforming the world with force (and ideas), whereas “paleo-cons” (e.g., Brzezinski and Mearsheimer) do not believe in transforming the world with force, not to mention ideas.3 Second, “paleo-cons,” or traditional realists, pay more attention to the danger posed by other great powers (i.e., Russia, China, and perhaps even Japan, India, and Germany) against the United States and less attention to dangers posed by “rogue states” and non-state actors (e.g., terrorists).

Finally, just because one believes in sustaining U.S. primacy does not automatically mean that one is a hawk or an offensive realist. Because realism believes in the utility of power in international politics, all realists inside the United States should want to sustain U.S. primacy. The differ-ence between offensive realists and defensive realists lies in how they want to sustain U.S. primacy.

Offensive realists will recommend the United States to pursue an offen-sive approach (e.g., launching preventive war if necessary), because offen-sive realists believe that only rigorously pursued offensive behaviors are capable of sustaining U.S. primacy. For offensive realists, the danger is doing too little, not doing too much (Wohlforth 1999, 39). Among offen-sive realists, the disagreement is not about the necessity of an offensive approach, but about how little is too little.

appendix ii / 191

In contrast, defensive realists will recommend the United States to adopt a strategy of self-restraint and moderation, because defensive realists believe that offensive behaviors are counterproductive for sustaining U.S. primacy. For defensive realists, the danger is usually doing too much, not doing too little (Art 2003; Ikenberry 1998; Walt 2005). Among defensive realists, the disagreement is not about the necessity of moderation and self-restraint, but about how much is too much.


Notes

Introduction

The (old) European Union, and to a less extent, South America, may have 1. evolved into a post-defensive realism or neoliberalism world, at least when it comes to how states within the community treat each other. I address this ques-tion in detail elsewhere.I have proposed a security strategy for China, in the spirit of defensive realism 2. (Tang 2003). I have also argued that China’s security strategy has decisively shifted from a more offensive realism type to a more defensive realism type since 1978 and that this shift partly accounts for China’s overall success in the past three decades (Tang 2008b).On “operational code,” see George (1969). See also chapter 4.3. For the problems with Mearsheimer’s framework, see Tang 2008a and the 4. ensuing discussion.See chapter 1 for a more detailed discussion.5.

One Clearing the Theoretical Underbrush

To some extent, this chapter is a partial response to Legro and Moravcsik’s 1. call for realists to redefine their paradigms more rigorously. Rosecrance (2001) voiced a similar but less strident complaint against the “degeneration” of real-ism. Legro and Moravcsik are guilty of defining realism narrowly and rigidly “so that realism can do little” (Feaver et al. 2000).The assumption that actors are strategic is more accommodating than the notion 2. that actors are rational, because “rationality” tends to have a heavy undertone of expected utility in the material sense. The assumption that actors are strate-gic can accommodate the possibility that actors may pursue some bad material goals or even nonmaterial goals (e.g., prestige, glory). As long as they pursue those goals with a strategic mind-set, they can be considered strategic actors. For a similar argument, see Waltz (1986, 330). As I shall show elsewhere, the assumption of strategic actors is indispensable for realism’s theorizing of actors’ behavior, and Waltz (1979) was wrong in doing away with this assumption.Some realists fail to emphasize this assumption (e.g., Schweller and Priess 1997; 3. Schweller and Wohlforth 2000), perhaps because they assume that power is all about material forces. But constructivists will not agree with this assump-tion. For many constructivists, in the words of Ernest Haas, “knowledge (i.e., ideas, ideologies, beliefs, etc.) is power.” For a recent nonrealist discussion on

194 / notes

power, see Barnett and Duvall (2005). To say that outcomes are determined . primarily by material forces, however, is not to argue that ideas do not play any role in international politics. The differences between realism and more ideationalistic theories (e.g., neoliberalism and constructivism) on the role of ideas in politics is not that realisms deny a role for ideas in shaping political behaviors, rather, realism takes material forces to be the ultimate determin-ing factor in shaping political outcomes, and makes ideas secondary factors (Schweller and Wohlforth 2000, esp. 99–101). Obviously, this assumption makes Keohane and Martin’s claim that institutionalism accepts realism’s core assumptions disingenuous and untenable. Nowhere in their understand-ing of realism’s core assumptions can one find material power (Keohane and Martin 2003, 73–74).For earlier discussions on the fundamental assumptions of political realism, 4. see Elman and Elman (1995, 183; 1997, 924); Frankel (1996, xiii–xviii); Gilpin (1996, 6–8); Legro and Moravcsik (1999, 12–18); Levy (1998, 145–146); Schweller (1997, 927); Schweller and Priess (1997, 4–6); and Schweller and Wohlforth (2000, 69–72).This point is inherently connected with (social) selection in the international 5. system (Waltz 1986, 330–331).The assumption that international politics is anarchical should be included 6. as an important core assumption, although it is as much an observation as an assumption.I believe that Taliaferro would agree with this interpretation, because he noted 7. that “neoclassical realists explicitly reject the injunction that theories ought not to include explanatory variables at different levels of analysis” (Taliaferro 2000–2001, 134). Levy (1998, 143–145) differentiated between employing the levels of analysis to refer to dependent variables and employing it to refer to independent variables.Waltz thought that he was explaining the recurrence of 8. de facto balance-of-power, but he was really explaining the recurrence of balancing behavior (Ruggie 1983, 267). Balancing is a type of state behavior while de facto balance of power is the outcome from the interactions between state behavior (i.e., balancing) and other factors (e.g., geography, power, and states’ ingenuity). Because there are other theories that are primarily driven by structural forces, I reserve neoreal-ism exclusively for Waltz’s approach and employ structural realism to denote all theories that are primarily driven by structural forces (e.g., Copeland 2000; Gilpin 1981; Mearsheimer 2001; Organski and Kugler 1980).Of course, even purely structural theories require some assumptions at lower 9. levels (e.g., units’ desire to survive) to operate. Hence, Powell (1994) ques-tioned whether it is even useful or possible to speak of structural theories (i.e., theories of international politics) in isolation.As Taliaferro (2000–2001, 134) pointed out, Rose (1998, 146) was mistaken 10. in taking neoclassical realism, offensive realism, and defensive realism as com-peting paradigms. The second image-reversed approach is more interested in bringing international variables to explain domestic politics (Gourevitch 1978). Although many (neoclassical) realists will readily admit a role for indi-viduals in international politics, they usually do not theorize factors at the individual level (Rose 1998, 156–157).

notes / 195

Because the three levels of analysis were not explicitly understood before 11. Waltz, classical and modern realists had not been conscious in grounding their explanations on different levels of analysis, and it is difficult to classify them into a particular level. I examine them in detail in chapter 6.Mearsheimer (2001, 19–22) called this strain of realism “human nature real-12. ism.” For two more recent discussions on these ultimate drivers of human behavior, see Lebow (2008) and Markey (1996).Thayer’s sociobiology-based offensive realism essentially seeks to explain 13. human beings’ supposedly inherent “lust” for power or “domination” with the biological roots of Homo sapiens. See Thayer (2004).Unfortunately, Waltz also asserted that only an explanation (of the repeated 14. recurrence of war in international politics) that is drawn from the third image can offer a “final explanation (because it does not hinge on accidental causes—irrationalities in men, defects in states)” (Waltz 1959, 231–232).Jack Snyder (1991) is generally credited for explicitly recognizing the diver-15. gence, although some important differences between the two realisms were already evident in Wolfers (1952). See chapter 6 for a more detailed discussion on Wolfers.For a more detailed discussion, see Tang 2008a and chapter 6 this volume. 16. For instance, Copeland’s theory of dynamic differentials and war, which is an offensive realism theory, is explicitly built to explain the probability of major war (Copeland 2000; see also Taliaferro 2001, 156–158). I thank Andy Kydd for discussion on this problem of possibility versus probability. Of course, these four components are not entirely separable (Jervis 1976, 48–49). An actor’s intentions are his preference over actions (for our purpose here, benign or malign) for reaching his goals. An actor’s interest (or goals, motives) drives his behaviors. Intentions are thus essentially “preferences over strategies,” whereas motives are “preferences over outcomes” (Powell 1994, 318–321). For structural realism, motives and intentions are independent of each other. Motive is structural in origin (i.e., dictated by anarchy), whereas intentions are driven at the unit level. Thus, while all states want power and security under anarchy, different states pursue these two goals with different strategies. For a useful but generally misleading discussion on motives and intentions, see Glaser (1992, 499–502). Earlier, Jervis (1976, 48–49) also defines intention as “actions” and makes it explicit that intentions are dif-ferent from “goals,” although he was mistaken in insisting that actions and intentions are identical.For instance, when facing an imminent threat, an offensive realist state will 17. seek cooperation (i.e., temporary alliance) with a third state even if it really wants to conquer the third state, see Mearsheimer (2001), 33. For the problem of trade-offs between long-term and short-term in Mearsheimer’s theory, see Lee (2002–2003).Because Taliaferro’s understanding about the differences between the two 18. realisms suffers from some serious errors, part of Taliaferro’s categorization of different realism theories into the two realisms has been misleading, if not wrong (see chapter 6 for details).In chapter 2, I show that Taliaferro’s stand is based on some mistaken under-19. standings about the security dilemma. Rose (1998) did not explicitly classify

196 / notes

Schweller as an offensive realist. In fact, Schweller (1993) had earlier argued that democracy should greatly mitigate the danger of war, which would put Schweller in the camp of “democratic peace.” Meanwhile, Taliaferro was not sure how to classify Kydd’s work even though Kydd (1997b), similar to Schweller, contended that the security dilemma could not possibly have been the major cause of war in history.I address the real problems associated with offensive realists’ deployment of 20. the security dilemma in Tang (2008a). Glaser (1997, 196) pointed out that Mearsheimer in his earlier writings either argued that the security dilemma does not exist or that it should never constrain states.In other words, Layne draws explicitly defensive lessons for states’ behaviors 21. from the thesis that seeking hegemony is counterproductive. Layne refuses to identify himself as either an offensive realist or a defensive realist, but he does note that when it comes to policy prescriptions for the United States, he has been consistently closer to Waltz than to Mearsheimer (Layne, personal com-munication, April 2007). Waltz, of course, has also consistently prescribed moderate policies for the United States (see chapter 6 for details).I address the offense-defense balance (ODB) as a key component of offense-22. defense theory (ODT) elsewhere (Tang 2010b). Briefly, I show that ODB is a theoretical hoax that holds little value for understanding international politics.Taliaferro (2000–2001, 160–161) then contradicts himself, noting, “Both the 23. offensive and defensive variants of neoclassical realism hold perceptions and misperceptions among top leaders may inhibit a state’s ability to respond to changes in relative capabilities.”For a similar stand, see Lobell (2002–2003, 166, 170). Taliaferro changed his 24. mind later (Taliaferro 2004, 233n21), without explicitly admitting his pre-vious mistaken stand on this point. Together, these two authors suggest that offensive realism is structural realism. Some critics of realism seem to sug-gest that whenever a theory admits domestic politics, it is no longer a realism theory (Legro and Moravcsik 1999). Most scholars (realists or not) deem such a position as a far too restricted vision of realism, see Feaver et al. (2000), and Jervis (1999).Glaser (1994–1995, 52–53) has been especially explicit in emphasizing the 25. structural nature of his “contingent realism.” The structural nature of Jervis’s theory is also evident (Jervis 1978, 167).D’Lugo and Rogowski (1992) advanced a similar idea in the context of the 26. Anglo-German naval race. This idea can be understood as one side of the Primat der Aussenpolitik (the primacy of foreign policy): In order to succeed externally, one has to do certain internal things properly (Zakaria 1992, 179).In chapter 4, I show that the fundamental difference on domestic politics 27. between the two realisms lies in what can be done about and through other states’ domestic politics.Elsewhere (Tang 2009c, 802), I emphasize that assuming that states seek 28. power or/and security is useful for theorizing state behaviors at the structural level, but of very limited value for understanding actual state behavior in the real word. Indeed, merely saying that states seek power/security robs much

notes / 197

explanatory power for state behaviors that is inherently possessed by states’ interests, an understanding clearly stated by Machiavelli’s truism of “ends justify means.” For understanding actual state behavior, we thus have to go down to states’ specific interests.Powell (1994, 317–321) first pointed out that structural realism does not 29. have a theory for preferences of outcomes—it “takes states’ preferences over outcomes as given”—although structural realism does seek to explain states’ preferences over actions.While Waltz classified armaments and developing cleaver strategies as inter-30. nal means of self-help, these two means are inherently tied to a state’s external means of self-help. How much armament is enough, what types of armament are best for security, and what kind of strategies are clever, are all inherently tied to the general external approach that a state adopts for its security. Hence, it is more appropriate to discuss them in conjunction with external means of self-help.I adopt Keohane’s definition of cooperation here. Cooperation is “a reaction 31. to conflict or potential conflict,” and it “reflects partially successful efforts to overcome conflict, real or potential” (Keohane 1984, 53–54). I discuss the problem of cooperation in greater detail in chapters 4 and 5.Thus, a major reason cooperation among states is difficult may not be because 32. states care about relative gains or fear being cheated, but rather because states are really aggressive and they simply do not want cooperation. See Tang 2008a and chapters 4 and 5 for a more detailed discussion.Scholars have mostly focused on gaining relative power through expansion or 33. conquest, but maximizing relative power can also be achieved by reducing the power of other states (e.g., sabotaging, weakening others’ alliance). I address this question in chapter 4.In this regard, Germany’s decision to annex Alsace-Lorraine after the 34. Franco-German War in 1870 is a classic case. While the annexation added to Germany’s relative power versus France, it made a Franco-German reconcilia-tion impossible, thus trapping Germany in an alliance with the weaker Austro-Hungarian Empire and reducing Germany’s relative power versus the alliance of Russia, France, and England. The net outcome was a reduction rather than an increase in Germany’s security (Trachtenberg 2003, 162–164).Influence can and should be understood as “relative power” because influence 35. is about relative power, and “power is a relative concept” (Hager and Lake 2000, 110; Jervis 1988, 334; Labs 1997, 11n33; Lynn-Jones 1998, 174). Zakaria’s replacing “relative power” with “influence” is a semantic exercise. After all, the title of his book is From Wealth to Power, not From Wealth to Influence.When commenting on Zakaria’s assertion that there is only a semantic differ-36. ence between the assumption that states maximize influence and the assump-tion that states maximize security (Zakaria 1992, 194), Lynn-Jones (1998, 177n47) pointed out: “This statement would only be true if influence-maxi-mizing is always the best strategy for maximizing security.”Because economic growth as an internal means of self-help can lead to both 37. absolute and relative power, a state can claim that it is not considering seeking relative power and seeking security as equivalent only if it does not see maxi-mizing relative power as the only sure external means of self-help.

198 / notes

Schweller (1996, 109–110) pointed out that this problem of relative versus 38. absolute gains in international cooperation is “artificial.” Exactly because the problem is artificial, however, Waltz’s inconsistency reveals his confusions more forcefully. A recent survey among IR scholars ranked this debate as the most unproductive debate in the past two decades (Peterson et al. 2005, 27–28).Waltz’s statement here is deceptive. By stating, “when faced with the possibil-39. ity of cooperation for mutual gain, states that feel insecure must ask how the gain will be divided” (Waltz 1979, 105; emphasis added), he seems to imply that some states (those that feel secure) will not be concerned with relative gains. Yet, because states can never feel completely secure under anarchy, his statement must mean that all states must always be concerned with relative gains (Schweller 1996, 102).Glaser (1994, 70–76) argued that the reason the concern for relative gain 40. does not always hinder cooperation is that a defensive realist state calculates gains in security rather than relative gains in power when deciding whether to cooperate, because security, not power, is the end for a defensive realist state. This reasoning is somewhat tautological. A better take might be that because a defensive realist state recognizes cooperation as a viable means of external self-help toward security, it does not always reject cooperation simply due to the concern for relative gains.This again reinforces the point that there cannot be a sound theory of state 41. behavior without looking at domestic politics. For similar arguments, see Rose (1998); Lobell, Ripsman, and Taliaferro (2009).For instance, the United States under Bill Clinton was defensive but became 42. more offensive under George W. Bush. See conclusion for a more detailed discussion.Jervis made both arguments briefly (Jervis 2001, 39–40). Thus, one of the 43. major weaknesses of the power transition literature has been that it tends to assume the hegemon to be always satisfied (and the rising power always dissatisfied), ignoring the possibility of “a disgruntled hegemon.” After all, power is not satisfaction. For a critique, see Resnick 2001, 554–555; Chan 2004, 210–211.Wolfers (1952, 484n3) and Jervis (2001, 40) reached the same conclusion 44. without developing their arguments fully. An illuminating illustration of the difficulties with this dichotomy is provided by the exchange between Schweller and Legro and Moravcsik on whether Schweller is a realist. See Schweller in Feaver et al. (2000, 176–177). For Schweller’s definition of stats quo versus revisionist states, see Schweller (1998, 24–25).By “other states,” I do not mean only other European states but former colonies 45. of the British Empire as well. Why would those former colonies accept that Britain was a status quo state back then? Hence, the dichotomy of revisionist states versus status quo states partly reflects the prevalent European/Anglo-Saxon ethnocentrism in IR. For European/Anglo-Saxon ethnocentrism in IR, see Acharya (2000); Buzan (1991, 299–300); and Levy (2003, 141–145).By implication, an imperialist power cannot possibly be a defensive realist 46. state. Indeed, Morgenthau labeled offensive realist states as imperialist states (Morgenthau 1948, 156).

notes / 199

Taliaferro (2004) extended this general concern for status and resistance to 47. “real or perceived” loss to arrive at a more general explanation for great power behaviors, especially their entanglements in the periphery, although his un-derstanding about the relationship between prospect theory and defensive realism is flawed (see introduction, this volume). Copeland’s theory of dy-namic differentials and preventive war essentially relies on a similar but nar-rower logic of loss aversion, even though Copeland dismisses prospect theory (Copeland 2000a, 2001).The second half of this argument is, of course, the typical argument for pre-48. ventive war, and preventive war is generally a signature of offensive realist states (see also Tang 2008a).Thus, it is at least inaccurate for Grieco to label realist states as “49. defensive positionalists” (Grieco 1993, 303). A more accurate interpretation will be that all realist states, both offensive and defensive, are “positionalists,” because they all believe that relative power is important (Schweller and Wohlforth 2000, 71–72), but only defensive realist states are “defensive positionalists” and offensive realist states are “offensive positionalists.” See also Schweller 1996, 109.Even such a qualified statement may not hold. Most students of IR certainly 50. would not consider those states that gained their independence from their former colonial masters after WWII as revisionist states or offensive realist states, even though those states unraveled the status quo at that time. Would anyone suggest that East Timor was a revisionist state, while Indonesia was a status quo state?In other words, fear and a sense of insecurity at least partially contributed 51. to their aggressive behavior. Here, I ignore the problem that even expan-sionist statesmen often justify their expansionist motives by saying they are seeking national security—adding this problem will only make the di-chotomy even more untenable. For a good discussion, see Zakaria (1998, 25–27).Zakaria’s dichotomy of maximal security-seeker versus minimal security-52. seeker has a similar undertone (Zakaria 1998, Chapter 2, esp. 28–31).Elsewhere, I argue that we should certainly admit that states’ different 53. goals often lead to different strategies if we allow more fine-grained differ-ences regarding goals, such as building empires or just defending its own territory.This part of offensive realism’s doctrine is often underappreciated in the 54. literature.Thus, balancing, especially balancing against threat (Walt 1987), is consis-55. tent with both offensive realism and defensive realism. See also Mearsheimer (2001), 44–48.Offensive realists will also note that offensive realist states had indeed found 56. ways to overcome balance and succeed in conquest in history. Otherwise, there would be no large states today. For a macro-history of human conquest and expansion, see Diamond (1997).Elsewhere (Tang 2008a), I show that non-offensive realism theories, in-57. cluding defensive realism, liberalism, neoliberalism, and constructivism, have much more in common than their proponents have admitted so far. As such,

200 / notes

we can speak of a dichotomy of offensive realist states versus non-offensive realist states. Since I will be dealing with realism in this book, however, I use the narrower dichotomy of offensive realist state versus defensive realist state throughout.When we say that a state is guided by offensive realism, defensive realism, or 58. some other grand theories of IR (e.g., liberalism), we are of course talking about approximation. Statesmen usually do not grasp the whole implications of these grand theories.This formulation is to accommodate the possibility that a defensive realist 59. state may be forced to take offensive actions when facing an obvious aggressor (i.e., an offensive realist state) that is going to launch an imminent attack. See the discussions in chapters 2 and 3.

Two The Security Dilemma: A Conceptual Analysis

At the very beginning, I differentiate the security dilemma from security 1. dilemma theory. The security dilemma is a concept for labeling a particular situation in international politics. Security dilemma theory is the body of the-ory that seeks to understand the underlying causes, regulations, and implica-tions of the security dilemma. See also section 5.Glaser’s version of structural realism is defensive realism or “contingent 2. realism.”The security dilemma requires lack of malign intentions to operate (see ensu- 3. ing discussion). As such, the security dilemma rejects assuming the worse over others’ intentions, and this in turn makes genuine cooperation, other than temporary alliance when facing a common opponent, possible. This is also why Jervis’s “Cooperation under the Security Dilemma” is the foundational work of defensive realism (Jervis 1978).I address the problems associated with applying the security dilemma as a 4. concept and a theory to ethnic conflict in Tang 2010a.The presence of disagreements on the security dilemma among scholars is 5. confirmed by a survey of prominent scholars that have either developed some aspects of the concept and theory or have utilized the concept as a major ana-lytical tool. A summary of the survey result is available on request.Glaser’s earlier effort fell short due to serious misunderstandings. Booth 6. and Wheeler’s more recent effort also fell short. See Glaser 1997; Booth and Wheeler 2008.While the dominant trend in the literature has been to stretch the concept of 7. security dilemma beyond recognition, there has also been a notable trend to strip down the security dilemma, especially to strip away the security dilemma of most of its material and psychological regulators, by the “rational choice” or “rationalist” approach toward (ethnic/civil) war. I address this literature in Tang (2010a).Butterfield and Herz wrote their treatises immediately after WWII, a war 8. that was initiated by a group of war criminals (Hitler, Mussolini, and Tojo). It is astonishing that both asserted that some of the greatest wars were not initiated by “criminals” or “the will to do deliberate harm,” but by a myste-rious entity called the security dilemma. Were they more concerned with the

notes / 201

looming cold war than with recent history? Butterfield (1951, 20) sounds so when he wrote that “the irreducible dilemma is at the basis of all the tensions of the present day.” Likewise, Herz (1966, 241) argued that “bipolarity gave the security dilemma its utmost poignancy.” This is an interesting question for a sociology of international relations theory.Butterfield’s “universal sin of humanity” can be reasonably linked with 9. Machiavelli’s ambizione, Reinhold Niebuhr’s “will to power,” and Hans Morgenthau’s “lust for power.”This notion can only be sustained by a biological or sociobiological under-10. standing about human nature. Yet, neither a biological nor a sociobiological understanding of human nature is tenable, and only a social evolutionary approach is valid. Note, however, only by attributing fear and thus the secu-rity dilemma to “universal sin of humanity” can Butterfield claim that the security dilemma is a critical driving force behind all human conflicts.This notion of “vicious cycle” will then be captured by Jervis’s “spiral.”11. After all, a dilemma is only meaningful if two sides end up in actual conflicts 12. even though neither side intentionally threatens each other.This uncertainty about each other’s present and future intentions under anar-13. chy is crucial to the formation and maintenance of the security dilemma. Yet, many authors tend to forget this point (e.g., Kydd 1997b; Schweller 1996), thus resulting in misunderstandings.Jervis, however, has not been consistent. For instance, he asserted that the 14. competition for colonies at the end of the nineteenth century was fueled by a security dilemma (Jervis 1976, 66), without recognizing that the competition for colonies was pursued by imperialists (offensive realist states) and there is no security dilemma among offensive realist states. See section 4.Again, Jervis has not been consistent. For example, when discussing the 15. Franco-German relationship before WWI, Jervis (1982, 361) argued that “the security dilemma here operated not as the unintended consequence of policy but rather as its object.”I believe that “material regulators” are a more fitting label than “structural 16. modifiers” or “physical modifiers.” Taliaferro (2000–2001, 137) recognized that structural modifiers are “material factors.”This uncertainty about the possibility that the other side may be a preda-17. tor may require the memory of past predators. See Schweller 1996, 91, 118; Copeland 2000, 25; and Kydd 2005, 18.From the very beginning of the advent of weapons (as tools of the early 18. humans), they have always been of dual purpose—to kill and to avoid being killed. Because our ancestors had to confront many powerful (nonhuman) predators before they began to confront each other, they must have accumu-lated some (offensive) power before the security dilemma between humans came into being.See also Waltz 1979, 186; and Snyder 1984, 462n2.19. In the context of ethnic conflicts, anarchy can be de facto anarchy (i.e. col-20. lapse of central authority). Here, anarchy should be understood as not the direct cause of the security dilemma, but rather as a necessary and permissive condition for the security dilemma to arise (Collins 2000, 12).

202 / notes

Although, logically, uncertainty over others’ intentions must precede fear and 21. they are ontologically separate, they are so closely intertwined that I list them together in figure 2.1. See also Tang (2008a).I am indebted to Stephen Van Evera for reminding me that the existence 22. of security dilemma only implies that some means of self-help will be self-defeating.The results from the brief survey are interesting here. With one dissenting 23. voice, all agree that anarchy is essential for a security dilemma. As a result, all agree that uncertainty about intentions and fear is also essential. On lack of malign intention, there is a significant lack of consensus. All those scholars who have applied the concept to situations in which one or both sides are malignant deny that this aspect is essential for a genuine security dilemma.Because the complete causal link from anarchy to the security dilemma and 24. then to war is rather lengthy, the thesis that anarchy automatically gener-ates the security dilemma and then conflict is open to question for many (e.g., Kydd 1997b, 191; Trachtenberg 2003, 174; Wendt 1992; Wagner 2003, 19–20). I address the link between the security dilemma and war in chapter 3.For instance, deterrence policies can also, and often do, produce unintended 25. and self-defeating consequences (Jervis 1976, 81, 90). In their new book, Booth and Wheeler (2008, 4–9) now called some of the undesirable outcomes from the security dilemma as “security paradoxes.” They, however, failed to emphasize that not all “paradoxes” are produced by the security dilemma.For instance, the early part of the cold war (circa 1944–1948) was a classic 26. spiral (i.e., the relationship between the Soviet Union and the United States steadily deteriorated), but this part of the cold war was not, nor caused by, a security dilemma, because the Soviet Union under Stalin was a malign state (see section 5 for details).This, of course, immediately begs the question how to read others’ intentions. 27. This problem is addressed in detail in chapter 5.Jervis (1976, 76n36, 76n37) did briefly allude to this issue in two foot-28. notes. Neither Butterfield nor Herz had said anything explicitly about the relationship.Some defensive realists were certainly guilty as charged. For instance, after 29. admitting that “European states adopted expansionist policies that fostered territorial conflict,” Stephen Van Evera then asserted that WWI “arose from a web of six misperceptions” and “that international cooperation may require controlling national misperceptions,” without stressing the genuine irrecon-cilable conflict of interest among the warring parties and the need to con-trol states’ expansionist ambitions in order to prevent conflict. See Van Evera (1985, 80–81, 117).This distinction is only superficially similar to Kenneth Boulding’s dichotomy 30. of real incompatibility and illusory incompatibility (Boulding 1959, 130). In reality, my formulation is very different from and subsumes Boulding’s di-chotomy. Most fundamentally, Boulding’s dichotomy is all subjective: How states define their security requirements determines whether their interests are compatible. As a result, Boulding’s dichotomy does not allow the possibility that even if there is subjective incompatibility, there may be no objective, or real

notes / 203

incompatibility. In contrast, my formulation allows many possible combina-tions of objective/subjective, real/illusory, compatibility/incompatibility. For earlier criticism of Boulding’s dichotomy, see Jervis (1976, 75–76).The situation in which there is neither objective nor subjective conflict of 31. interest between two states is of little interest to students of international pol-itics. The world is in perfect harmony, and the challenge for states is how to coordinate for collective benefits. There will be few, if any, active security dilemma in this world.Apparently, this subjective irreconcilability of interest is a form of “illusory 32. incompatibility” (Boulding 1959, 130).Thus, it is only under this scenario that defensive realism will rely mostly on 33. misperceptions to drive states into actual conflict.For instance, although both sides are benign and willing to settle for a com-34. promise, one or two sides try to drive a hard bargain, and the other side may well end up in believing that the state is malignant, thus significantly worsen-ing the security dilemma.An example of the first possibility is that both sides want to use a water sup-35. ply source, but mistakenly believe there is not enough water for both of them even though the supply could meet both sides’ needs through a cooperative water sharing and management scheme. An example of the second possibil-ity is that, under de facto anarchy, two ethnic groups that previously lived together peacefully end up in believing that their security depends on exclud-ing or subjugating each other, due to the fear of being dominated by the other group.An example of this scenario will be that a chauvinistic (majority) group 36. demands that other (minority) groups be subjugated to the dominance by the former (e.g., the Croats versus the Serbs in Croatia after the breakup of the former Yugoslavia).This scenario is likely to be rare and unlikely to last long because states tend 37. to pay more attention to the potential evil intention of others and it is difficult for a true aggressor to conceal its true intention well and for long (see chapter 5). A situation in which both sides are already malignant yet both sides also believe that their interests are compatible would be even rarer.A typical situation under this scenario will be that one state wants to grab a 38. piece of territory from the other state, or both states want the same piece of territory, and both sides understand the situation.Schweller (1996) was thus correct that a real security dilemma depends on the 39. “status quo,” or more precisely the defensive bias, of defensive realist states. He, however, is wrong to take Waltz as the quintessential defensive realist. Waltz has not been consistent in his elaborations, although he seems to be more in the defensive realism camp (see chapters 1 and 6 for details).This immediately raises the empirical question, “How common are the secu-40. rity dilemmas?” Elsewhere (Tang 2009b), I argue that this question can only be resolved via a social evolutionary reading into the history of international politics. Briefly, security dilemmas were rare in the pre-1945 world, but have become far more prevalent in the post-1945 world. See also chapter 3 in this volume.

204 / notes

I specifically argue that competitive self-help anarchy allows for the rise of 41. the security dilemma because there have been other types of anarchies. By arguing that defensive realism does rely on structure to drive conflicts, I do not necessarily agree that structure alone can drive international conflicts. Indeed, I shall argue that anarchy alone cannot dictate war. The meaning of anarchy for state behavior and war has not been properly and adequately un-derstood. For earlier but underdeveloped elaborations, see Copeland (2000, 24–25); Kydd (1997b, 128); Wendt (1992; 1999).In maintaining that defensive realism relies on misperceptions to drive con-42. flicts in international politics, Schweller further accused defensive realism of believing that conflicts and conflicts of interest among states are the result of misunderstanding and misperceptions, thus violating one of the fundamental assumptions of political realism toward international politics—that interna-tional politics is fundamentally conflictual, and conflicts of interest among states are genuine. See earlier discussions.A more precise formulation should be, “One state’s security can make others 43. feel less secure.” See also Waltz 1979, 108; Mearsheimer 2001, 32–33; and Tang 2008a.Jervis recognized some of the potential drawbacks of this dichotomy, but did 44. not elaborate (Jervis 1976, 75–76).For a similar stand when discussing ethnic conflict, see Rose 2000, 378, 45. 380.Boulding (1959, 130) recognized the two sides of the security interest, as 46. Wolfers (1952) did earlier. Unfortunately, Boulding chose to focus exclu-sively on the subjective side. Discussions after Boulding have generally failed to appreciate that the security interest has two sides and the implications of that for the security dilemma. As a result, these discussions have also focused exclusively on the subjective side of security interests (e.g., Collins 2000, Jervis 1976; 1978; Roe 2000; 2001; 2004).Of course, if the illusory incompatibility is due to one or both sides’ mispercep-47. tion that their security interests are incompatible, yet neither side harbors malign intentions toward each other, the situation is a classic security dilemma.Indeed, if the security dilemma cannot drive states to severe illusory incom-48. patibility, it will have little role in driving states to actual conflict.Of course, such a stand means that security-seeking motive does not neces-49. sarily equate to benign intention. Glaser, however, is not so sure of this “inde-pendence” of motive and intention because he also noted that “they may share a single source” (1992, 502n6). I partly agree with Glaser that motive (that is, goals) and intention are somewhat independent of each other.Indeed, this is the position Glaser took more recently, emphasizing that even 50. if a state is greedy, as long as the motive for their behavior is security-seeking, the security dilemma still operates, although in a somewhat weaker fashion (Glaser 1997, 174, 190–191, 199–200).This, of course, contradicts Glaser’s initial position that motive and intention 51. are independent of each other. Glaser seems to have been struggling with himself on this issue.Indeed, some offensive realists argue explicitly that their theory starts with the 52. assumption that states seek security, and it just happens that their theory leads

notes / 205

to the conclusion that maximizing relative power, which mandates offensive or intentionally threatening behavior, is the best and the only sure means to-ward security (Labs 1997, 4–5, 11; Mearsheimer 2001, 29, 34).In other words, their feeling of insecurity at least partially contributed to their 53. aggressive behavior.Note the similarity between the problems associated with security-seeking 54. motive and the problems associated with (illusory) incompatibility.Of course, this dependence on others does not necessarily make Japan inse-55. cure objectively (it makes it vulnerable), as the remarkable economic recovery of Japan after WWII demonstrated.The four expansions examined below can be categorized into two strains. 56. The first strain—represented by Snyder, Wheeler and Booth, and Collins—essentially expands the concept to cover situations that simply cannot be accommodated by the concept or to cover situations that have already been covered by it. The second strain, represented by Glaser, operates at a more fundamental level of the concept. Because of the centrality of the security dilemma in defensive realism, critics of defensive realism have also tried to undermine the logical foundation of security dilemma theory. Those mis-guided expansions of the security dilemma theory certainly did not help the critics either. Some of the criticisms of the security dilemma theory were actu-ally directed against some of the skewed understandings of the concept (e.g., Kydd 1997b, 126, commenting on Glaser 1994, 67–70).“Deadlock” captures the situation between two offensive realist states, and 57. Jack Snyder was correct in not labeling it as a sort of (security) dilemma.Judging from the verbs that Snyder employed to describe the effect of per-58. ceptual factors on the security dilemma, it seems that he understood clearly that perceptual factors can only regulate the severity of the security dilemma but cannot give birth to a new kind of security dilemma. The verbs employed by Snyder include “tighten” (161), “exacerbate or alleviate” (162), “intensify” (164), and “feed” (169). Snyder, however, was sometimes sloppy with his wording. For instance, he used “create and intensify” when it should be just “intensify” when discussing the regulation of the security dilemma by mili-tary technology and geographical circumstances (157).To their credit, Booth and Wheeler (2008) have now dropped this misleading 59. label. Because their new definition of the concept does not contain the lack of malign intentions, however, their new definition still allows “deliberate secu-rity dilemma.” See earlier discussions.Glaser’s elaboration did generate some important insights. However, most of 60. these insights are applicable to the broader phenomenon of the “spiral,” but not “the security dilemma,” and he apparently takes the security dilemma and the spiral model as equivalent. See section 4.Glaser actually contradicted himself, because elsewhere he repeatedly empha-61. sized that uncertainty is critical for security dilemma and that insecurity is predicted by the structure of international politics (Glaser1992, 507; 1997, 189–193, 195).I keep the “spiral model” because it has gained wide acceptance. Reiter (1995, 62. 8n5) implies that the spiral model is different from security dilemma theory. See also Collins (2000, 175, 192n51).

206 / notes

Indeed, Jervis never defined the spiral and the spiral model, nor did he define 63. the security dilemma rigorously.There are two other possibilities: one believes that a spiral and a security di-64. lemma are different, or cannot be sure of the relationship between the two. All three positions are present in the brief survey that I conducted.American Heritage Dictionary65. (New York: Dell Publishing, 1983), 658; Concise Oxford Dictionary, 10th ed. (Oxford: Oxford University Press, 1999), 1384; McMillan English Dictionary for Advanced Learners (London: McMillan, 2002), 1377.This differentiating also allows us to resolve the critical question of 66. whether the security dilemma has been a major cause of war in history. See chapter 3.One may argue that these imperialist states only threatened each other on 67. interests beyond survival and security (i.e., colonies). Yet, exactly because they were willing to risk war on nonsecurity interests, they were offensive realist states.Thus, Jervis actually alluded to Snyder’s ill-informed labels of “imperialist se-68. curity dilemma” and “deadlock” without labeling them. The principal cause behind Jervis’s slip was that he often left the concept under-specified. Another possible cause may be that Jervis too readily accepted statesmen’s justification of their expansive behaviors on face value (Snyder 1991, 12).Snyder further noted that one needs a calibrated combination of both de-69. terrence and assurance to handle the situation (Snyder 1985, 179; see also Christensen 2002, 8–10). I develop this argument further in chapter 4.Snyder and Jervis (1999, 19) noted the possibility of a spectrum or continuum. 70. Their formulation, however, starts from an ill-informed angle because they believe that malign intention is compatible with the security dilemma.Jervis (2001) prefers “deep” while Snyder (1985) prefers “vicious” security 71. dilemma.Thus, there is a gray area between a “deep” security dilemma and a spiral, and 72. this may make it difficult to differentiate between the two in practice.Both “expansionist threat” and “mutual threat” can also be due to the pres-73. ence of malign intention at the very beginning. If this is the case, then the situation should not be confused with a deep security-dilemma-derived situa-tion. A deep security-dilemma-derived situation is an outcome of the security dilemma dynamics, whereas an “expansionist threat” and “deadlock” due to the presence of malign intention at the very beginning is not.Because regulators do not require a lack of malign intention to operate, I do not 74. differentiate the security dilemma from the spiral in this part of the discussion.The four material regulators are geography, polarity (i.e., structure), military 75. technology (i.e., objective offense-defense balance), and the distinguishability of offensive and defensive weapons.Monica Duffy Toft (2003) singled out the concentration of minority groups 76. as another material regulator of the ethnic spiral or security dilemma. Because concentration of minority groups is essentially an unmixing of minority groups, concentration and mixing are two sides of the same coin. I thus group the two as one variable.

notes / 207

Indeed, much of the security dilemma literature implicitly assumes a symmetric 77. distribution of power between the two actors (e.g., Snyder 1984, 462n2).Collins’s discussion on the interaction between China and the Southeast 78. Asian states after the cold war is a rare good treatment of this subject (Collins 2000, Chapter 5). For a general treatment on the impact of asymmetric dis-tribution of power on state relations, see Womack (2006).The delicate triangular relationship between the United States, the Southeast 79. Asian nations, and China is one good example of this situation (Collins 2000, Chapter 5).Unfortunately, in the voluminous discussion of managing the U.S.-China 80. relationship, few have paid serious attention to the possibility of a security dilemma or spiral between the two states. For exceptions, see Christensen (1999, 2000); Copeland (2000, esp. 52–54); Johnston (2002); Tang (2003, Chapter 9); Zhang and Tang (2005).For instance, although a hegemon may not face an objective (i.e., real) secu-81. rity threat from the rising power, the hegemon may still worry about the insecurity of its hegemonic position. The recent outcry in the United States against China’s growing economic and political presence in Latin America and Africa reflects this oversensitivity. Some Chinese analysts have been aware of Washington’s oversensitivity and its implications for China’s foreign policy (Tang 2003, chap. 9; Zhang and Tang 2005).Much of the existing literature on hegemonic stability and hegemonic war 82. implicitly and explicitly assumes that the hegemon must necessarily be a satisfied power (or status quo power, or defensive realist state). But this is not necessarily the case, since “power is no satisfaction” (Chan 2004; Jervis 2006). Of course, if both the rising power and the hegemon are offensive realist states, the situation becomes almost impossible to manage, and actual conflict may well become inevitable.For instance, during the rampage against ethnic Chinese in Indonesia 83. in 1998, the Chinese leadership faced extreme public pressure to act more forcefully. On the subject of sustained discrimination and crime against the Chinese diasporas in Indonesia, see the Minority at Risk (MAR) project Web site, http://www.cidcm.umd.edu/inscr/mar/chronology.asp?groupId=85003 (accessed June 26, 2006).Only one psychological regulator of the security dilemma—the perceived 84. (subjective) offense-defense balance—has received sustained attention from IR scholars (Christensen and Snyder 1990; Glaser 1992; Jervis 1976, 1978; Levy 1984; Snyder 1991; Van Evera 1999). I address the problem of offense-defense balance in Tang (2010b)Although Van Evera (1994) did not explicitly employ the security dilemma to 85. link nationalism with war, many of the dynamics identified by Van Evera can and perhaps usually do work through the security dilemma or spiral mecha-nism to drive states and ethnic groups into conflicts.One can certainly argue that concern for reputation among U.S. leaders dur-86. ing the cold war played an important role in driving the spiral between the United States and the former Soviet Union. For the concern for reputation and U.S. behavior, see McMahon (1991). Regarding analogous thinking,

208 / notes

when a state considers any compromise as appeasement, it is impossible to unwind a security dilemma.For two good reviews of prospect theory in IR, see Levy (1997) and Mercer 87. (2005).This aspect also ties in to asymmetric power. The Dniestr region in Moldova 88. may be one such case. Traditionally, the Russians there enjoyed privileged positions. After the collapse of the Soviet Union, however, these privileged positions were in serious jeopardy. And this fear of “replacement effect” among the Russian elite in the Dniestr region may have been the major reason that propelled them into seeking eventual secession (Kaufman 1996a, 1996b). For “replacement effect,” see Acemoglu and Robinson (2006).I thank Li Mingjiang for bringing this case to my attention.89. Many may counter that democratic peace is a liberal theory, and thus it cannot 90. be mixed in with the realism literature on war and peace. But if democratic peace theory is to study war and peace (which has long been the exclusive domain of realism) and realism is also paying growing attention to domestic politics, I do not see any reason democratic peace cannot be jointly discussed with realism theories, including the security dilemma theory. For a similar call, see Copeland (2000, 7–9).

Three The Security Dilemma and War Revisited

For simplicity, we use “states” to denote all independent political entities (e.g., 1. tribes, chiefdoms, states, empires, and warlords) in history.Indeed, other than Butterfield, the only realist who had explicitly argued that 2. the security dilemma has been a major cause of war has been Mearsheimer (2001), the quintessential offensive realist. I have addressed Mearsheimer’s hijacking of the security dilemma in detail in Tang 2008a.This means that Glaser must conclude that the security dilemma has been a 3. major cause of all wars, if he pushes his argument to its logical conclusion.This statement can only be partly right. While it may be reasonable to argue 4. that “most wars originate in the structure of the international political sys-tem,” to deny that Hitler or Napoleon alone can cause war is simply untenable (Byman and Pollack 2001). Waltz certainly did not really offer a neorealism theory of the origins of war per se, contra to his claim (1988, 620).I address the problems of ODT in detail in Tang (2010b). 5. The question why preemptive or preventive war is the only route through 6. which the security dilemma causes war is addressed below.China’s entering the Korea War came close to be a war mostly caused by a 7. security dilemma. Both the United States (but not McArthur) and China wanted to avoid fighting each other, yet still ended up at war.These cases included the Peloponnesian War, the War of Spanish Succession 8. (the French-Habsburg War), the Seven Years War, and the Napoleonic Wars. In addition, Copeland examined the second Punic War (Carthage-Rome), Louis XIV, WWI, WWII, and the cold war. I am not suggesting that Copeland’s interpretation must necessarily be right. I am merely suggesting that we need to take a pause when two fundamentally different theories take the same set of cases as supporting evidences.

notes / 209

Of course, Van Evera had no intention to prove the thesis that the security 9. dilemma has been a major cause of war, although he may believe in it.Levy noted the criticalness of WWI for the spiral model and security-dilemma 10. theory (Levy 1990–1991, 189n3), without differentiating the spiral model and the security dilemma theory. For a recent exchange on the causes of WWI as a critical case, see Lieber (2007) and Snyder and Lieber (2008).Despite this, it is difficult to classify Kydd as an offensive realist, as Taliaferro 11. (2000–2001, 146) noted. See also chapters 5 and 6. James Fearon (1995) also casts doubt on the thesis that the security dilemma has been a major cause of war. Unfortunately, Fearon has stripped the security dilemma down to incomplete information and commitment of problem (which is just another label for uncertainty about intentions). On Fearon’s stripping-down of the security dilemma, see Tang (2010a).Glenn Snyder actually pointed out that Schweller neglected the critical role 12. of uncertainty in giving birth to the security dilemma when commenting on Schweller, but Schweller (1996, 119n96) decided to ignore Snyder’s point. Somewhat similarly, Spirtas (1996, 410–412) pointed out that Schweller was overconfident in states’ ability to correctly perceive each other’s intentions.Kydd (1997b) was ambivalent and inconsistent on the question whether the 13. security dilemma will ever cause war. In one place, he stated categorically that “the search for security does not lead to conflict, in spite of uncertainty about motivations (152).” In another place, however, he also stated that “this type of uncertainty is usually overcome (117),” thus implying that it may not be overcome.Essentially, Kydd (1997a, 2005) takes fear as a function of uncertainty, and it 14. does not have a life of its own. Kydd’s mistake is similar to that of Schweller. Schweller called fear as “misplaced,” while Kydd called it “mistaken” and also pretty much ignored it altogether. Kydd’s negligence of fear might be due to the fact that game theoretical models have yet to be able to model fear.Indeed, it is not easy to identify a situation as a security dilemma, even with 15. hindsight (Jervis 2001, 37–42). Kydd noted that a more powerful security-seeking state may nonetheless have less incentive to signal its benign inten-tions (Kydd 1997a, 395), thus implying whether a state adopts reassuring measures or signals benign intentions is conditional. I discuss reassurance in detail in chapter 5.In other words, it may be difficult to unwind a spiral generated by a security 16. dilemma under certain circumstances in which many physical and perceptual modifiers of the security dilemma work together to drive up the spiral (Glaser 1992; 1997, 198; Snyder 1985, 154–155). Kydd’s model, sophisticated as it is, could not possibly model all the material and psychological modifiers of the security dilemma.For instance, when Schweller (1996, 103–104) pointed out that status quo 17. states will always be able to signal their motivations when defense is domi-nant, he also implied that it would be difficult for them to do so when offense is dominant (Schweller’s argument is based on ODT). Likewise, when Kydd noted that “there are equilibria in which security seekers can avoid war” (Kydd 1997a, 373; emphasis original), he also implies there may be equilibria in which states cannot avoid war. See also Kydd (1997b, 117; 2000a, 326).

210 / notes

In fact, Kydd’s examination of the trust-building process between Gorbachev’s 18. Soviet Union and Regan’s America showed just how difficult it is to convince one’s adversary that one’s intentions are genuinely benign (Kydd 2000a, 340–351. See also Larson 1997, Chapter 6). Kydd’s focus on democracy’s ability to signal benign intentions surely does not help the generality of his argument either. See chapter 5 below.As I argue in chapter 5, this fourth cause has not been a major cause behind 19. the lack of signaling benign intentions.For earlier elaborations of the differences of the two types of war, see Levy 20. (1987); Reiter (1995, 6–7); Snyder (1985, 160–161); Van Evera (1999, 40–41, 41n18).In other words, these two types of war can overlap. In addition, because pre-21. ventive wars are about long-term threats, given a time horizon that is long enough, it would be possible to classify many wars as preventive wars. This is a serious methodological problem that the literature has yet to address. Reiter is wrong in insisting that “the essence of preemption is that it is motivated by fear, not by greed” (Reiter 1995, 6), thus implying that fear has less a role in driving preventive war. In reality, fear also plays a prominent role in pre-ventive war, partly because fear is universal under anarchy. Sometimes, the label of “inadvertent war,” understood as “unintended war” or “unintentional war,” is employed to label “preemptive war.” “Inadvertent war” should not be understood as “accidental war” (i.e., a war triggered purely by an accident). Moreover, it is difficult to argue that a war can be caused purely by an acci-dent (i.e., the assassination of Archduke Francis Ferdinand and his wife alone could not have caused WWI), because the decision to mobilize for war has to be a political decision. To avoid confusion, I avoid “inadvertent war,” because even preemptive war is no accident.This means that Reiter’s cases potentially cover preemptive wars induced both 22. by nonsecurity dilemma spiral and by genuine security dilemma. Thus, if he finds that preemptive wars have been rare even when including wars induced by nonsecurity dilemma spirals, then preemptive wars induced by genu-ine security dilemma will be even rarer. Taliaferro (2004, 296n45) rejected Reiter’s thesis without offering any counter evidences.Reiter identifies a war as preemptive war only if preemption is the major 23. motive behind the war.“Security-driven expansion” is a type of preventive war because it aims to 24. prevent other states from having the possibility of threatening one’s security by weakening, if not eliminating, those states first. Interestingly, Copeland’s theory of dynamic differentials theory and preventive war, an offensive real-ism theory that Taliaferro wrongly classified as a defensive realism theory, made the most explicit case that the security dilemma leads to preventive war (Copeland 2000, esp. Chapter 2).Thus, Leffler (2005) argued that there are actually much more continuity 25. between the preventive “Bush doctrine” and traditional U.S. foreign policy. For arguments that the Bush doctrine is revolutionary, see Jervis (2006); Nelson (2005); Offner (2005).This long or very long time horizon subsumes the problem of shifting inten-26. tions by others (i.e., turning from benign to malign). Because decision-makers

notes / 211

can be wrong, real sign of relative decline is not absolutely necessary. Relative decline, of course, can be caused by one state’s own stagnation and/or other states’ more rapid growth. See, Copeland (2000); Gilpin (1981); Levy (1987); Modelski (1978); and Organski and Kugler (1981).Thus, examining cases in an offensive realism world (i.e., pre-1945), both 27. Lebow (1981, 254–263) and Jervis (1989b, 153–164) found that many deci-sion-makers believed this “inevitability of war.”Obviously, when the target state of a state’s preventive war is really aggressive 28. at the moment when the state decides to launch the preventive war, the situa-tion is not a security dilemma either.See also Jervis (1976, 63). Jervis (1978, 186) listed remarks by one of Catherine 29. the Great’s ministers, “that which stops growing begins to rot,” and Japanese decision-makers’ belief that the alternative to increasing Japan’s dominance in Asia was to sacrifice Japan’s very existence as illustrations of the mentality of status quo states, without realizing that Tsar Russia and Imperial Japan were both offensive realist states, not status quo states, and there would be no secu-rity dilemma between any one of them and any other state, or between the two themselves. The two remarks cited by Jervis reflect the prevalent Social Darwinism-Geopolitics thinking at that time, and both Social Darwinism and Geopolitics were thoughts of offensive realism (see chapter 6 this volume). The fundamental cause behind Jervis’s erroneous stand has been that he did not differentiate the security dilemma and a spiral (see chapter 2 this volume).Hence, resorting to preventive or preemptive war must also mean that a 30. state discounts the (future) utility of alliances (Snyder 1985, 160–161). Such a stand is again more consistent with offensive realism’s operational code (Mearsheimer 2001, 29). See also Lee (2002–2003). Many, based on ODT, may argue that another reason (defensive realist) states do not launch preven-tive wars may be that defense usually holds advantage over offense (Snyder 1985, 158; Van Evera 1999). As I argue elsewhere (Tang 2010b), however, this possibility, although valid, is not as robust as many proponents of ODT have asserted. For much of human history, conquest was frequent and often successful (Tang 2009b).In contrast, trying to forestall a war with other means may not only avoid 31. the war, but also provide a state with the time to learn another state’s true (benign) intentions (see chapter 5 this volume).Thus, while Lynn-Jones (1998, 176, 181) called for a theory of security-driven 32. expansion within the defensive realism camp, a theory of security-driven expansion will be more an offensive realism theory, not a defensive realism theory. Taliaferro (2004) attempted to develop a defensive realism theory of imperial expansion, but his theory was really an offensive realism theory (see introduction).Abraham Lincoln, replying to the argument of his law partner William 33. Herndon that President Polk had invaded Mexico for a defensive purpose, put it pointedly: “Allow the President to invade a neighboring nation when-ever he shall deem it necessary to repel an invasion . . . and you allow him to make war at pleasure. If today he should choose to say he thinks it necessary to invade Canada, to prevent the British from invading us, how could you stop him? You may say to him, ‘I see no probability of the British invading

212 / notes

us’; but he will say to you, ‘Be silent; I see it, if you don’t’ ” (Quoted in Brodie 1973, 257).As this point, it is difficult to tell whether the war is preventive or preemptive, 34. or both.Jack Snyder (1991, 21–26) listed two additional conditions for expansion: 35. resources of conquered territories can easily be exploited, and multipolarity. The first condition is subsumed under offense advantage (Van Evera 1999). Copeland discussed the impact of polarity upon states’ motive to launch preventive wars (Copeland 2000a). These material factors, however, do not touch upon intentions.I elaborate on the social evolutionary approach toward social changes (in-36. cluding changes in international politics) in detail elsewhere. For an applica-tion of the approach at the systemic level, see Tang (2009b). For an application of the approach at the state level, see Tang (2008b).Wendt’s earlier insight that the anarchy of the competitive self-help kind is 37. an institution or “what the states make of it” has the same logic behind it. Anarchy only plays a permissive role for the coming of the competitive self-help anarchy, and there is no inevitable link between anarchy and competitive self-help (Wendt 1992). Competitive self-help, of course, is the perquisite of the existence of the security dilemma. In sum, just as there is no inevita-ble link from anarchy to war although anarchy is “routinely cited as a root cause and explanation for the recurrence of war” (Fearon 1995, 384; see also Trachtenberg 2003, 174), there is no inevitable link from anarchy to the secu-rity dilemma (and then to war).Hence, it was wrong for Posen (1984, 69) to assert that “status quo policies are 38. the rule rather than the exception”. For most of our history, status quo policies and status quo states were the exception, not the rule. Offensive realism poli-cies were the rules (Tang 2009b).I offer two timelines for the coming of security dilemma as a widespread prop-39. erty of international politics: Westphalia in 1648 (an optimistic take) and the end of WWII in 1945 (a conservative take). See Tang (2009b) for details.As Robert Art (1973, 8) aptly put it, “Each wished to be ‘top dog’ and that nei-40. ther could tolerate the other’s being so.” Many have pointed out Wilhelminian Germany’s hegemonic ambition in leading to WWI, but failed to appreci-ate the fact that Britain also harbored hegemonic ambition or even offensive intentions (versus Germany) at that time. For exceptions, see Chan (2004) and Trachtenberg (1991).For the latest manifestation of this tied hands, see Jack Snyder’s defense of 41. defensive realism against Lieber’s frontal assault (Snyder and Lieber 2008).Both defensive realism and offensive realism mistakenly seek to be a 42. timeless theory (Tang 2009b).Dan Reiter (1995, 33) made a similar point when examining preemptive war 43. as a central prediction of the spiral model and ODT.

Four The Operational Code of Defensive Realism

Again, an “operational code” is “a set of precepts or maxims of tactics and 1. strategy that are derived from a relatively coherent belief system” (George

notes / 213

1969). I am thus interested in explicating what a state should do strategically if it is a defensive realist state (or an offensive realist state) and, by implication, I am also suggesting that a state will be punished if it violates the operational code that it should obey. The second half of this statement is crucial for un-derstanding the power of realism and other grand theories of international politics. I discuss this assumption and the power of realism elsewhere.In other words, offensive realism has a very simple strategic calculus. Not 2. surprisingly, the minimal realism advocated by Legro and Moravcsik (1999) looks strikingly similar to offensive realism.Mearsheimer (2001, Chapter 5) developed a useful, but somewhat less well 3. articulated scheme. His scheme suffers from the same problem as that of Schweller (see ensuing text).When a weak state faces a powerful threat without readily available allies, 4. appeasement (and/or bandwagoning for security) is a rational choice (Schweller 1994), and it is consistent with prescriptions by both offensive realism and defensive realism.My definition of engagement is thus substantially different from that of 5. Resnick (2001, 559–561), which focuses exclusively on positive sanctions without emphasizing the hedging element.As such, at least in principle, a deterrence strategy should not be an indepen- 6. dent grand strategy either. A deterrence strategy should be an integral part of engagement or containment (passive and active).This is most evident in the often heated debate on its China policy inside the 7. U.S. policy and academic circle. Few proponents of engagement with China would identify engagement as a purely reassuring policy toward China. Most proponents of the engagement policy understand that engagement with China is conditional, although they differ on the exact conditionality of such a pol-icy. Indeed, hedging has become the new buzzword in the debate on China policy (e.g., Medeiros 2005–2006).Gaddis (1981, 79–83) had a different but somewhat overlapping distinc- 8. tion of symmetrical and asymmetrical containment. I believe that passive and active containment is a more suitable dichotomy. A mostly passive con-tainment approach was perhaps what George Kennan had in mind, and this explains why he came to disagree with the more active containment approach favored by Paul Nitze.This is partly due to the two schemes’ offensive realism bias. See chapter 6 9. and Tang (2008a).Here, it is implicitly assumed that the invitation to cooperate from the defen-10. sive realist state is a genuine offer rather than a publicity stunt or a trap. See chapter 5 for details.In other words, whereas extensive cooperation between two states requires a 11. significant amount of mutual trust, reassurance or extending an invitation of cooperation in order to gauge the other side’s intentions does not require the presence of extensive trust. See chapter 5 for details.Buckpassing or hiding generally implies that the first state has already taken 12. the second state as a (potential) threat, although the former does not want to contribute to the actual effort of balancing against the latter. Buckpassing can also be an offensive tactic against a third party (i.e., the buck-catcher).

214 / notes

For a good discussion, see Mearsheimer 2001, 157–162. As such, buckpassing or hiding is a less friendly strategy toward the second state than explicit neutrality.Please note that for peace to hold here, I mean de facto balance-of-power, 13. rather than balancing-of-power (i.e., balancing behavior). For the distinction of the two, see Ruggie 1983, 267.Obviously, some of them do interact and overlap with each other. For 14. instance, when a state achieves regional hegemony, it also achieves an over-whelmingly robust deterrence posture with regard to all other states in the regional system.Glaser’s conclusion on this point is similar, though somewhat weaker: “a 15. greedy state will find that an arms buildup is its best option under a wider range of conditions.”In light of this, Elizabeth Kier’s culturalist explanation of France’s and 16. Britain’s defensive military doctrines before WWII is deeply flawed because she failed to differentiate security strategy from military doctrines and to understand that military doctrines follow from security strategy (Kier 1997). As a result, she failed to understand that because France and Britain before WWII were more defensive than Nazi Germany, the two states rationally adopted a defensive strategy and a defensive military doctrine. In contrast, Hitler, with an expansionist goal, had to adopt an offensive strategy and an offensive military doctrine (Porch 2000, 167–169).The primary example of offensive alliance is the Axis powers in WWII.17. When a weak state faces a powerful aggressor and potential allies against the 18. aggressor are not available, bandwagoning for protection is perhaps better understood as a defensive measure, and it is consistent with both offensive realism and defensive realism.In ancient China, the Kingdom of Wu paid dearly for not totally annihilating 19. its once vanquished victim, the Kingdom of Yue. Yue eventually came back and eliminated Wu (Hui 2005, 58–60).This is perhaps the only occasion that Waltz emphasized the benefit of offen-20. sive behaviors in his work. See chapter 6 for a more detailed discussion on Waltz.This strategy is perhaps the most cynical and cleverest of all the indirect strat-21. egies. The Kingdom of Yue successfully employed this strategy against its archenemy the Kingdom of Wu (Hui 2005, 59–60). Brzezinski’s strategy to induce the Soviet Union to invade Afghanistan reflects a similar logic. See “Interview with Zbigniew Brzezinski, How the US provoked the Soviet Union into invading Afghanistan and starting the whole mess,” Le Nouvel Observateur, Jan 15–21, 1998, on line at http://members.aol.com/Bblum6/brz.htm (accessed September 27, 2006). See also Gaddis 1981, 96.Of course, the supposedly general preference for offensive doctrines among 22. military organizations may complicate this problem (Posen 1984).It is thus profoundly misleading for defensive realists to assume that com-23. mon interest among states is universal or unconditional and that it has been anarchy and/or the security dilemma per se that has prevented states from cooperating with each other and reaping the common gains (e.g., Jervis 1978, 66; 1982, 360; 1988, 319). The presence of common interest is conditional,

notes / 215

not universal or unconditional, under anarchy. For a more detailed discussion, see Tang 2008a.Unfortunately, states may tend to focus on the second dimension while giving 24. little thought to the first dimension. They simply assume that all conflicts of interest are genuine. Such a mentality unnecessarily exacerbates the security dilemma/spiral and heightens the tension between them.It will become clear that these two questions are also inherently tied to dif-25. ferentiate states into offensive realist states and defensive realist states. See section 4 and chapter 5 for details.This, of course, implies that even a genuine defensive realist state may occa-26. sionally commit the sin of being opportunistically greedy. In other words, between an ideal offensive realist state and an ideal defensive realist state, there is a continuum of different types of states (see chapter 1). On how to differentiate offensive realist states from defensive realist states, see section 4 and chapter 5.The Oye-edited volume contains two brief empirical studies on security coop-27. eration, by Jervis (1986) and Van Evera (1986). Unfortunately, Van Evera’s study is deeply flawed because he attributed the root cause of WWI almost exclusively to misperceptions. Jervis provided several hypotheses on the for-mation and durability of concert of great powers as a type of security regime.For instance, while Ross (1995) detailed how the United States and China 28. negotiated cooperation and Schultz (2005) formalized this process of mak-ing peace during the cold war, the cooperation between the United States and China during the cold war was mostly an alliance strategy when facing a common threat rather than a genuine strategy of cooperation between the two major foes in the conflict. Fortunately or unfortunately, one can make the case that some of the empirical studies of cooperation with a minimalist neoliberalism approach actually fall within defensive realism (see chapter 6).For a dated survey on CBMs, see Desjardins (1996). I discuss CBMs in greater 29. detail in chapter 5.A neoliberalist state will behave very much like a defensive realist state 30. when it comes to war and peace, only more optimistic and institution-con-scious. I thus treat neoliberalist states and defensive realist states as essentially alike when discussing making security strategies.This formulation accommodates the fact that some offensive realism theories 31. draw diametrically opposite (i.e., offensive) implications for state behaviors, despite admitting the prevalence of balancing behavior and the existence of the security dilemma (e.g., Copeland 2000; Mearsheimer 2001). Of course, one should not expect statesmen to grasp all the implications of the two phe-nomena, as even scholars have misunderstood the security dilemma and its implications one way or another (chapter 2).For instance, immediately after WWII, when the United States was still not 32. too sure about Stalin’s intentions, the United States extended various invi-tations for genuine cooperation to Stalin’s Soviet Union (e.g., putting the atomic weapon under some sort of international control). Strikingly, the rationales behind many U.S. decision-makers’ decision to support such co-operative measures were exactly to gauge Stalin’s true intention (Kydd 2005, Chapter 4). In a personal communication to the author (Aug, 2006), Jervis

216 / notes

also emphasized that statesmen (and scholars) generally employ this folk wis-dom to gauge other states’ intentions.This means that it is more difficult to differentiate between an offensive real-33. ist state and a defensive realist state when both are facing another offensive realist state.Hawks, however, got the picture only partly right. They neglected the 34. possibility that an offensive realist state and a defensive realist state can even run an arms race differently.This does not exclude the possibility that a weak (defensive realist or offensive 35. realist) state may bandwagon with a very powerful aggressor when allies are unavailable (Schweller, 1994, 94–95).Thus, Jervis exaggerated somewhat when he asserted that the security di-36. lemma (actually a spiral) would be so vicious that defensive realist states (i.e., status quo powers for Jervis) “must act like aggressors” (Jervis 1978, 187; em-phasis added). Under most circumstances, a defensive realist state can behave differently from an offensive realist state, if it chooses to.This requires that military postures can be reasonably differentiated into 37. offensive and defensive types. I address this issue in detail in Tang 2010b.In contrast, offensive realism believes that even the spiral model and the de-38. terrence model combined would be flawed. States should go for active con-tainment, if not preventive war (Fischer 1996, 275).Kydd (2000b) failed to recognize that even an arms race can be pursued dif-39. ferently, thus neglecting the possibility that arms race can serve not only as a device for signaling resolve to continue the arms race, but also as a device for signaling reassurance.Exercising self-restraint and being willing to be constrained are the two sides 40. of the same coin, because being willing to be restrained is a form of self-restraint. When a state accepts constraints over its behavior even if it has the power to overpower those constraints, it is exercising self-restraint.Because most militaries have multiple tasks to shoulder or multiple contin-41. gencies to be coped with, it is critical for two defensive realist states to take these cooperative measures in order to avoid considering their armaments as targeting each other and to contain the security dilemma between them when they have to develop military capabilities for other possible scenarios (or adversaries).A recent prominent case has been the confidence-building process between 42. Russia and China (1988–1995) and, to a less extent, between India and China (1989–1998). See chapter 5.Of course, it is often tricky and difficult to support the moderates within the 43. adversary’s camp while maintaining a robust deterrence and defense posture. This is the classic dilemma or tension between the deterrence approach and the spiral approach (Jervis 1976, Chapter 3), but this is the dilemma that a state has to face “if security policy is to be expedient” (Wolfers 1952, 497). Defensive realism does not advocate for political changes (i.e., through wars) in another country. The peace-making approach of defensive realism should not be confused with the liberal theory of “democratic peace.” The defensive realist approach does not emphasize the connection between regime types and

notes / 217

moderation. After all, Gorbachev came to power in a totalitarian system (and George Bush came to power in a democratic system).In other words, extensive cooperation will inevitably evolve into some kind 44. of institutional arrangement. As such, institutions do have a role to play in defensive realism (see chapter 6).Such sophistication does not automatically give defensive realism an edge 45. when it comes to solving practical problems in the real world. It depends on the nature of the world (Tang 2009b).

Five Reassurance: A Defensive Realism Theory of Cooperation-building

Elsewhere (Tang 2008c), I show that the notion that intentions are difficult 1. to gauge is underspecified, and thus misleading. To properly understand the problem of uncertainty about others’ intentions, we need to unpack intentions into two broad types: malign intentions and benign intentions. Moreover, the two general types of intention pose almost diametrically opposite challenges for our perception. Malign intentions are easy to detect and difficult to hide whereas benign intentions are too easily dismissed and discounted thus “hard to detect.” This general fact is part of the fundamental reason international coop-eration is difficult while noncooperation easy, and the ultimate cause behind this fact lies in our evolutionary past. During the long evolutionary process leading to modern humans, we have been strongly conditioned to detect and overreact to potential danger posed by others—be it human or other nonhu-man animal species—while ignoring and underreacting to potential friendli-ness from others, for the sake of our survival.In this sense, offensive realist states actually act according to certainty over 2. others’ intentions! See also Wheeler and Booth (2008, 34–38). For a more detailed discussion on Mearsheimer’s stand on uncertainty over others’ inten-tions and the security dilemma, see Tang (2008a).I avoid using “building peace” because peace can be either “cold” (i.e., a lack 3. of war) or “warm” (i.e., a cooperative relationship). Building extensive coop-eration is equivalent to building “warm peace.” For “cold peace” and “warm peace,” see Miller (2000).Paul Midford (2002) and Michael Glosny (unpublished Ph.D. dissertation, 4. MIT, 2008) have argued that reassurance can also serve the purpose of pre-venting oneself from being balanced, in the case of Japan and China respec-tively. Similar arguments can be found in Ikenberry (2000), Mastanduno (1997), and Walt (2005). These authors all start from a balance-of-threat perspective (Walt 1987). I shall argue that preventing oneself from being bal-anced can also be a means toward building more cooperation.I do not explicitly deal with the possibility that others’ intentions can 5. change—a possibility that has been repeatedly emphasized in realism litera-ture (e.g., Jervis 1976, 62; Copeland 2000b, 199–206). I address this problem of anticipating others’ future intentions separately elsewhere.For instance, although I do mention the impact of having common allies or 6. common opponents on reassurance, I have essentially treated reassurance as if it is a purely dyadic phenomenon.

218 / notes

Much confusion has resulted from not explicitly differentiating these four 7. factors. For earlier and incomplete discussions, see Huth (1997, 75–78); Tang (2005, 39–40); Tang (2008, 453–455). I address the cognitive challenges for perceiving these factors from a social evolutionary perspective in Tang (2009d).For structural realism, motives and intentions are independent of each other: 8. motive is structure-driven (i.e., dictated by anarchy), whereas intentions are unit-level driven. This is incorrect: ends certainly dictate means, at least to some extent. For a useful but generally misleading discussion on motives and intentions, see Glaser (1992, esp. 499–502). For a more detailed critique of structural realism’s position, see Tang (2009d).Mistrust and distrust are almost exact synonyms, and many have used them 9. interchangeably or prefer one over the other (e.g., Larson 1997). Hardin (2002) prefers distrust whereas Kydd (2005) prefers mistrust.Fear of others is thus only partially underpinned by our uncertainty over oth-10. ers’ intentions, thus mistrust or lack of trust of others (Tang 2008, 451–453). Apparently, I am talking about “specific trust” rather than “generalized trust” here. For the differentiation of specific and general trust, see Hardin (2006). My definition of trust and mistrust thus follows Kydd (2005, 6–12). See also Deutsch (1958, 265–267); Gambetta (1988, 217–219); Larson (1997, 19–21). I refrain from discussing trustworthiness, which can be understood as one’s reputation that he/she can be trusted. Both Hardin (1996, 28–29) and Kydd (2005, 42n6) noted that many have conflated trust and trustworthiness.Collins Cobuild English Dictionary11. (London: Harper Collins, 1995), 1373.David Edelstein’s (2002) formulation that cooperation is a means to induce 12. benign intention (presumably from malign intention) is deceptively similar to the formulation here. His formulation is fundamentally wrong (see section 2).These earlier discussion of cooperation do not explicitly connect cooperation 13. with trust.Many do not differentiate cooperation from cooperative strategies. Others 14. have differentiated the two terms only implicitly (e.g., Milner 1992, 467–470; Glaser 1994–1995; Jervis 1999).The challenge is, of course, how to know whether a (reassurance) signal is a 15. signal for genuinely benign intention rather than a trap (with malign inten-tion behind it) or a ploy. Although I do not address this issue in length here, the logic and the means for facing this challenge can be found in section 4. I thank Robert Jervis and David Rousseau for reminding me of the possibility of false reassurance and these classic cases. For a more in-depth discussion on why states resort to false reassurance attempt or “counterfeit diplomacy,” see Montgomery (2008).Even a verbal reassurance signal carries some risk because it can be attacked by 16. one’s domestic audience as appeasement, thus risking losing some of one’s do-mestic stand, especially if the other side does not reciprocate (see section 3).I thank an anonymous reviewer for suggesting this formulation.17. Early discussions on the problem of credibility in deterrence signals thus suf-18. fered from erroneously taking credibility and reputation (for resolve) as non-relational (e.g., Schelling 1966).

notes / 219

Apparently, this equation means that we tend to underestimate the credibility 19. of a reassurance signal. The actual equation that decision-makers and, quite often scholars too, deploy to assess the credibility of a reassurance signal is not the equation above, but an equation that further underestimates the credibil-ity of a reassurance signal. The ultimate cause behind this again lies in our evolutionary past (Tang 2008c).Here, the attempt must be a tangible concession (i.e., attempt in deeds, see 20. section 4) rather than merely rhetoric, because only tangible concessions can be taken advantage of.For earlier and incomplete discussions, see Mitchell 1991, 410; Weinberger 21. 2002, 94.As Yamagishi et al. (2005, 276) noted, this conventional wisdom is perhaps 22. mostly caused by the prevalence of deploying (reiterated) prisoner’s dilemma (PD) game as the major tool for studying cooperation and trust for much of the time. In PD game, one cannot signal trust without full cooperation and one cannot signal distrust without full defection. See also Kydd (2005, 9–11). For the influence of PD game in studies of international cooperation, see Jervis (1988). The notion that “conflict requires some mistrust ex ante” is correct. When two parties trust each other (i.e., they are certain that they are both benign) completely, no conflict is possible.Although Montgomery (2006, 153n9) claimed to use “motives” to 23. denote preferences over goals, he was not entirely consistent. Here, “motives” should be “intentions” (i.e., preferences over strategies).Aaron Hoffman (2007) emphasizes an institutional approach and challenges 24. an incremental approach toward trust-building and cooperation-building. But his challenge suffers from three critical errors. First, he fails to recognize that institutions (i.e., rules and norms) for enforcing cooperation are results of cooperation ex ante. Second, he identifies the European Union at the “Treaty of Rome” in 1986 as a low-trusting cooperation relationship! Third, he greatly underestimates the impact of an external enforcer of cooperation (often the United States).Yamagishi et al. (2005) labeled extensive cooperation as “trustful 25. cooperation.”Hence, these two authors’ logic often self-contradicts. For instance, after 26. asserting that “trust . . . is usually a necessary (though not a sufficient) con-dition for states to cooperate,” Larson then noted that “trust might then be a useful by-product, not a precondition for mutual cooperation” (1997, 6, 10).Admittedly, Glaser (1992; 1994–1995) emphasized the differentiation 27. of motives (or preferences over goals, i.e., power or security) from inten-tions and that motives are the key to differentiate different types of states. Unfortunately, as shown in chapter 1, Glaser’s stand is invalid. More impor-tantly, when Glaser talked about benign or malign “motives” (1994, 67–70), he was really talking about intentions.Indeed, as stated in chapter 4, even when facing a not-so-incorrigible offen-28. sive realist state, a defensive realist state still has every incentive to reassure the offensive realist state as part of its broader strategy of engagement. This is so because only engagement can turn the offensive realist state into a defen-sive realist state, thus enhancing the security of the defensive realist state. Of

220 / notes

course, changing an offensive realist state will not be easy, and the defensive realist state may well forsake the strategy of engagement due to fear and/or the mutual hostility generated from the spiral dynamics.Both Daniel Treisman’s “rational appeasement” and Taylor Fravel’s the-29. sis that regime insecurity leads to cooperative behavior operate on this logic. See Treisman 2004; Fravel 2004–2005.Edelstein did not define and differentiate cooperation and cooperative strat-30. egy. His thesis will make a bit more sense if he talks about engagement as a cooperative strategy.Such a stand, which Edelstein claims constituting “a challenge to offen-31. sive realism theory and a needed improvement in defensive realism theory” (38), is only marginally different from offensive realism’s position. The only meaningful difference between Edelstein’s stand and the standard offensive realism stand may be that whereas offensive realism believes that states are inherently aggressive, Edelstein (13) argues that “intentions can change.”The other crucial case in Edelstein’s analysis is Munich. I address Edelstein’s 32. mistaken interpretation of the cause of the Munich tragedy elsewhere.In other words, only the subjective balance directly influences state behavior, 33. whereas the objective balance influences the outcomes of the interaction of states’ behavior.Lynn-Jones (1995, 681n58) asserted that “the historical record reveals 34. many examples of leaders making judgments about the offense-defense balance” by citing Snyder, Howard, Van Evera, and Christensen and Snyder. Yet, all these authors employed WWI as their principal case!Military posture includes operational doctrines (e.g., operational tactics 35. and operational rules of engagement), pattern of troop deployment (i.e., for-ward or non-forward deployment), and deployment of weapons (Porch 2000, 179).Montgomery (2006, 158) then contradicts himself immediately: “Because 36. security seekers are concerned foremost with their continued survival and because they fear that other states are greedy and prefer to expand at their expense, only signals that clearly diminish a state’s ability to do so [i.e., to expand] will differentiate the two types of actors.”The proper formulation on this problem should read as follows. Because an 37. aggressive state will be reluctant to sacrifice concrete offensive military advan-tages in an uncertain effort to lull the adversary, a nonaggressive state must go beyond tokens and make concessions weighty enough in offensive capabilities so that a state contemplating attacking or coercion would be unwilling to make them. Such concessions, however, do not have to risk defensive capabil-ities. They only have to be weighty enough in reducing offensive capabilities to be differentiated from the behaviors of an offensive realist state, because an offensive realist state will be unwilling to give up potential offensive advan-tages that can be derived from any relative gain in offensive capabilities. Such a formulation makes it explicit that reducing some offensive capabilities is not only reassuring but can also be relatively safe.Indeed, focusing on undifferentiated total power may prevent certain reassur-38. ance attempts from bearing fruit. When the receiver of a reassurance attempt that focuses on relative power demands the attempt’s initiator to reduce its

notes / 221

relative power, the initiator is less likely to obey. Such a dynamics inevitably makes reassurance attempt more difficult to succeed.The problem of cost and credibility for signals of resolve in conflict may be 39. quite different from the problem of cost and credibility for signals of benign intentions toward cooperation.Khrushchev did interpret Kennedy’s speech as a reassuring signal (Larson 40. 1997, 28, 242).By implication, then, if a state does not back its words with deeds, it will be 41. deemed to be less trustworthy.While neoliberal institutionalism has traditionally dominated the discussion 42. on this topic, defensive realism has an equally, if not more, valid claim to this domain. Jervis (1976, 67), for instance, pointed out long ago that “states must employ and develop ingenuity, trust, and institutions” to manage their secu-rity under the security dilemma. In fact, many countries might have jumped onto the bandwagon of institutions mostly on the logic of defensive realism.Glaser (1992, esp. 526–533; 1994–1995, esp. 67–70) dealt with reassur-43. ance involving military deeds in detail. Unfortunately, his discussion suffers from some important errors. First, he heavily relied on ODT, especially ODB, and this in turn makes many of his discussion suspicious (see previous discus-sion). Second, his three major categories of measures (i.e., defense emphasis/unilateral defense, arms control, and unilateral restraint) often overlap with each other. Third, while recognizing that unilateral self-restraint is almost always reassuring (see ensuing text), his formulation on this aspect is incor-rect. Glaser insists that unilateral restraint means “reducing military capabil-ity below the level it [a state] believes would otherwise be necessary for deterrence and defense” (1994–1995, 69, emphasis added). Such a formulation violates realism’s assumption that states are strategic actors. The correct formulation should be that a benign state should reduce its offensive military capability below what is required for aggression.In contrast, supporting insurgency within the other side is usually taken as a 44. very provocative posture. A similar measure is “competitive decolonization,” as noted by Hager and Lake (2002).In contrast, trying to profit from another state’s window of vulnerability will 45. be almost surely taken as an aggressive gesture by that state. Among allies, if such a scenario develops, the situation becomes much worse. Because the vul-nerable state actually hopes (or expects) a helping hand from an ally, an ally that seeks to profit from the state’s window of vulnerability is most likely to be deemed as “stabbing in the back” by the vulnerable state.In contrast, expanding war aims is usually a signature of an offensive realist 46. state (Labs 1997).The logic here thus corroborates Jervis’s earlier insight that the more secure 47. a state feels, the more likely it is likely to cooperate (Jervis 1978). While an extremely fearful defensive realist state will be reluctant to initiate reassur-ance just like a malign state, the former will still behave differently from the latter when both are offered an opportunity to reciprocate to a reassurance program.These factors, however, do not influence states’ intentions. They merely influ-48. ence states’ determination (i.e., resolve) for cooperation and conflict.

222 / notes

Under most circumstances, it is easier to assess the success of a particular reas-49. surance attempt than to assess the success of a particular reassurance program simply because a particular reassurance program may take months, years, and even decades to reap its final fruit.Unfortunately, many have explicitly or implicitly neglected this possibility, 50. although their discussion often implied such a possibility (e.g., Mitchell 1991; Montgomery 2006).This part of the discussion draws freely from many existing contribu-51. tions, in particular Osgood (1962, esp. Chapter 5); Etizoni (1962, 92–110); Jervis (1970; 1976); Mitchell (1991); Glaser (1992; 1994–1995); Larson (1997); and Kydd (2005, Chapters 2 and 7).Indeed, if a high ranking official in your government is openly against 52. your reassurance program, you should reprimand or even fire him/her to impress the other side about your sincere wish for cooperation. For exam-ple, because Dulles was so vocally against Eisenhower’s effort to reciprocate Khrushchev’s reassurance attempts, Eisenhower could have fired Dulles to impress Khrushchev.Hence, Tit for Tat is not a good tactic of reassurance, and Osgood’s GRIT 53. is definitely better. For a good discussion, see Stein 1991, 52–56. See also Keohane 1986, 10–12.Unfortunately, exactly because you have designed the reassurance signal to 54. test the other side’s intention, you may well prematurely conclude that the other side is malignant if they did not respond to your signal positively. This might have been a major reason many reassurance programs have ended in failures. I thank Robert Jervis for emphasizing this important point.In contrast, doves who have made the initial conciliatory moves usually 55. pay an electoral price if those moves are not reciprocated by the other state (Colaresi 2004).CBMs generally exclude reassurance with words. Because reassurance covers 56. reassurance with words, reassurance subsumes CBMs. Theorists on reassur-ance have generally ignored CBMs, thus neglecting a critical source of empir-ical evidence, to the detriment of theory-building.

Six Recategorizing Realism Theories

For instance, Taliaferro classified offense-defense theory (ODT) as a defen- 1. sive realism theory. This is simplistic and thus misleading. ODT is actually very complex. See Tang (2010c)A rigorous categorization of different theories in the field not only sharpens 2. the debate among different theories and makes implicit assumptions and logic more explicit, but also demands theorists to be more explicit, precise, and consistent with their assumption, deduction, and policy implications. Plus, a rigorous categorization of different theories in a field normally requires an ex-tensive and in-depth understanding about the field, which some intellectual historians may not possess.Glaser’s position is agnostic or somewhere between the position of “no role” 3. and the position of “no essential role”: “Contingent realism [i.e., defensive

notes / 223

realism] leaves open the question of whether institutions will play a role in making cooperation possible (Glaser 1994–1995, 84).”Institutions are simply embodiments of ideas such as rules and norms (Boland 4. 1979).This realism institutional approach certainly did not fool Men Honghua, a 5. Chinese analyst. For him, international institutions are simply “the wings of U.S. hegemony” (Men 2005). Of course, an institutionalist America will per-haps be more acceptable to the world than a unilateralist one.Most institutionalism theories focus on state behavior, although some institu- 6. tionalism theories do imply that ideas determine outcomes from the interac-tion of state behaviors. Realism also assumes that international outcomes are determined primarily by material power (e.g., Jervis 1982, 358).Mitzen (2006) seems to head in such a direction, although she focuses on 7. conflict.I avoid “idealism” because it has another meaning (i.e., utopianism). 8. Only under this light can we understand why Keohane and Martin’s asser- 9. tion that institutitonalism accepts realism’s core assumptions is so disin-genuous. Power, not to mention material power, was nowhere to be found among the realism assumptions listed by them (Keohane and Martin 2003, 73–74)!This is so despite that Schweller self-identified his thesis as an offensive real-10. ism argument.Offensive realist states do form offensive alliances that are explicitly 11. geared for conquest and expansion (e.g., the Axis powers), and offensive alliance might have been a major cause of war. IR theorists have generally neglected offensive alliances, making only passing references to it in their lit-erature (e.g., Jervis 1978, 204–205; Schweller 1994).Modelski’s long-cycle theory and war also falls into this category (Modelski 12. 1978; 1987). Gilpin (1981, 94–95) explicitly stated that a state’s intention is exclusively dependent upon its capability. His stand is almost the same as made by Mearsheimer (2001, 37, 45).Thucydides’s often-quoted passage, that the rise of Athens as a power and the 13. fear it caused in Sparta made the Peloponnesian War inevitable, conveys the same thesis. From a strategic perspective, the Thucydides-Copeland thesis is far more plausible than the Gilpin-Organski-Kugler thesis. As Copeland noted, “There is no logical reason why a state should attack while it is still rising, since by simply waiting, the state will be able to achieve its objec-tives more easily, and at less cost” (Copeland 2000, 13; see also Chan 2004). The Gilpin-Organski-Kugler thesis may again reflect a strong Anglo-Saxon status quo and ethnocentric bias. For earlier discussions about the preventive motive, see Levy (1987).Under this light, Zakaria (1992) criticized Snyder not for attempting to ex-14. plain imperialism, but for offering a “wrong” explanation for imperialism. Whereas Snyder believes that conquest and expansion are generally counter-productive and has to be explained by domestic pathologies, Zakaria believes that conquest and expansion can be productive and it does not have to be the result of domestic pathologies.

224 / notes

Christensen (1996) also examined how Mao in China initiated the second 15. Taiwan crisis in 1958 to mobilize the nation for his eventually disastrous Great Leap Forward. Because Mao’s mobilization was geared toward domestic agen-das, however, this case falls outside of the realm of international politics.Because this literature implicitly assumes that the other side is inherently 16. aggressive and is generally uninterested in resolving the conflict through coop-eration, an argument can be made that it is more akin to offensive realism.For other criticisms of Waltz’s inconsistencies, see Labs (1997, 8), Wendt 17. (1999, 104–107, 249); Zakaria (1992, 194n43); Zakaria (1998, 29–30). For conflicting interpretations of where Waltz stands between offensive realism and defensive realism, see Kydd (2005, 14n14); Snyder (1991, 12n36). Indeed, Kydd’s two interpretations of Waltz on the same page contradict each other.Thus, it is not surprising that Barry Posen, Stephen Walt, and Stephen 18. Van Evera, all of them Waltz’s former students, are also defensive realists (Mearsheimer 2006, 109).Elsewhere, I shall argue that Carr, Niebuhr, Morgenthau, Wolfers, Kissinger, 19. and Kennan, by partly focusing on the problem of morality in politics, repre-sented a transition period that eventually led to the clear bifurcation within the realism camp. For a similar stand on Niebuhr and Morgenthau, see Craig (2003); Lebow (2003). I shall rank Thucydides and Morgenthau as the two most complex thinkers. I do not discuss Hobbes because I think Hobbes was mostly concerned with domestic politics.Thucydides was perhaps the first realist, but his 20. History contains both offen-sive realism and defensive realism components, making it difficult to catego-rize. See Welch 2003.It is now generally agreed that Machiavelli believed that the republic is a bet-21. ter political system for waging war (e.g., Discourse, 131). Machiavelli advo-cated for a republic strictly on instrumental grounds. Thus, The Prince and Discourses represent the same teaching, and together they form a complete theory of power politics, both domestic and international. For similar inter-pretations, see Fischer 1995–1996; Forde 1992, 373, 378, 387; and Markey 1996, 144–147.In other words, domestic politics should be explicitly geared toward provid-22. ing the material base for external expansion. A similar theme emerged in Kautilya’s Arthasatra and Machiavelli’s Discourses. For Shang Yang’s role in engineering Qin’s rise to universal domination, see Hui 2005, Chapter 2.Han Fei Zi,23. with almost 90, 000 Chinese characters, is a long treatise even by today’s standard. Han Fei Tzu was translated into English in 1924. Shang Jun Shu was translated in 1928. Both translations, however, remain obscure.Of course, this does not mean that contemporary offensive realism is a direct 24. descendent of geopolitics and social Darwinism. For an earlier discussion on social Darwinism and geopolitics that put it in the larger context of the devel-opment of realism, see Schweller 1996, 96–98.Geopolitics without social Darwinism retains a following among many prac-25. titioners (e.g., Brzezinski 1998), and to a less extent among theorists too.This trade-off was then captured by Jervis (1976) as the choice between the 26. spiral model and the deterrence model. Offensive realism does away with the trade-off (see chapter 4 this volume).

notes / 225

This point is akin to arguing that offensive realist states tend to seek absolute 27. security (chapter 1).A prominent example has been Jack Snyder. See earlier discussion.28.

In Lieu of Conclusion: Policy Implications

In this sense, I am not a defensive realist. I believe defensive realism, although 1. important, is inadequate for understanding the whole complexity of interna-tional politics through history.This is also due to the fact that malignant intentions are difficult to hide. Due 2. to this evolutionary social psychology of fear and trust (Tang 2008c), states are more likely to attribute malignant intentions to others. As such, in today’s world, the more likely danger is not a failure to detect another state’s malign intention but rather a tendency to attribute malign intention to another state and then pursue an unwarranted confrontational policy.Here, it is important to differentiate institutions (as rules, norms, or regimes) 3. from organizations, which are enforcers of rules and are underpinned by rules at the same time. See Duffield (2007).Happily, the election of Barack Obama as president in 2008 may have turned 4. the United States into a defensive realist state.Patrick E. Tyler, “U.S. Strategy Plan Calls for Insuring no Rivals Develop,” 5. New York Times, March 8, 1992, A1. For the story behind this doctrine, see Mann (2004, 198–215). It is important to note that defending primacy as a goal is consistent with both offensive realism and defensive realism. Thus, although John Ikenberry (2000), Stephen Walt (2004), and Michael Mastanduno (1997) all wanted to preserve the unipolar moment, they have been vocal critics of the Bush doctrine and their policy prescriptions are more consistent with defensive realism. What makes the 1992 DPG an offensive realism doctrine is that it calls for the United States not merely to defend its primacy but to actively prevent or block the emergence of any potential com-petitor to American power. The paper trail of the 1992 doctrine is now avail-able at: http://www.gwu.edu/~nsarchiv/nukevault/ebb245/index.htm.In 1992, U.S. defense spending was larger than the combined spending of the 6. next fifteen countries. By 2008, U.S. defense spending, excluding its spend-ing in Afghanistan and Iraq, has become larger than the combined spend-ing of the next twenty-five countries. Author’s calculation from Stockholm International Peace Research Institute dataset (www.sipri.org).These authors generally label the doctrine as “imperialistic,” “revolutionary,” 7. or “a doctrine of transformation.” While there have been intense debates on whether the United States is an empire now, most observers have no quarrel that the behaviors of the United States have become (more) imperialistic after September 11, if not earlier. For debates, see, “ISP Forum: American Empire,” International Studies Perspectives 9 (3), August 2008.Of course, when the United States acts in the interest of itself and others, 8. other countries should lend their support. And in return, these great powers should explicitly demand reciprocity from the United States when they need the support of the United States. Through this interactive process, certain cooperative code of conduct may be forged.

226 / notes

Appendix I: World War I and the Cold War Revisited

Another prominent thesis about the causes of WWI, Van Evera’s thesis that 1. WWI was caused by a “cult of the offensive” (Van Evera 1999, Chapter 6), will be examined in Tang (2010b).Of course, it is debatable whether Khrushchev really understood the cause of 2. the cold war: Did he think that Stalin merely pursued expansion in a wrong way (i.e., Stalin was too heavy-handed), or did he think that Stalin should not pursue expansion at all? The first scenario will not make Khrushchev a defensive realist, but the second scenario will (Jervis 2001, 43). Here, I am taking the position that Khrushchev was essentially a defensive realist. As such, I agree with Larson that there might have been a missed opportunity between Khrushchev and Eisenhower (Larson 1997).Because Gorbachev’s effort was so extraordinary, however, one can reasonably 3. argue that the cold war ended with total capitulation by one side rather than by genuine cooperation between the two sides. Gorbachev basically submitted to most U.S. demands, and the United States merely accepted Gorbachev’s genu-ine offer without offering much in return (Schweller and Wohlforth 2000).

Appendix II: Liberals, Conservatives, Doves, and Hawks

I thank Emrys Chu for this catchy label of “Limps.”1. For some examples, see the quotations in Walt 2005, 62–63, 72.2. This is one of the main reasons “paleo-cons” can join with their defensive coun-3. terparts (i.e., defensive realists) in strongly opposing the invasion of Iraq (e.g., Mearsheimer and Walt 2003). “War with Iraq is not in America’s National Interest,” New York Times, September 26, 2002.

Bibliogr aphy

Acharya, Amitav. 2000. “Ethnocentrism and Emancipatory IR Theory.” In (Dis)Placing Security: Critical Evaluation of the Boundaries of Security Theory, edited by Samantha Arnold and J. Marsha Beier. Toronto: Center for International and Security Studies.

Akerman, Alice. 1994. “Reconciliation as Peace-building Process in Postwar Europe.” Peace and Change 3: 229–250.

Arrow, Kenneth J., et al. (eds.). 1995. Barriers to Conflict Resolution. New York: W.W. Norton.

Art, Robert J. 1973. “The Influence of Foreign Policy on Seapower: New Weapons and Weltpolitik in Wilhelminian Germany.” Sage Professional Papers in H&H Studies 2: 9–19.

Axelrod, Robert. 1984. Evolution of Cooperation. New York: Basic Books.Axelrod, Robert, and Robert Keohane. 1985. “Achieving Cooperation under

Anarchy: Strategies and Institutions.” World Politics 38: 226–254.Ba, Alice D. 2003. “China and ASEAN: Renavigating Relations for a 21st Century

Asia.” Asian Survey 4: 622–647.Bacevich Andrew. 2002. American Empire: The Realities and Consequences of US

Diplomacy. Cambridge, MA: Harvard University Press.Baldwin, David. 1993. Neorealism and Neoliberalism: The Contemporary Debate.

New York: Columbia University Press.———. 1997. “The Concept of Security.” Review of International Studies 1: 5–26.Barnett, Michael and Raymond Duvall. 2005. “Power in International Politics.”

International Organization 59 (4): 39–75.Betts, Richard. 1999. “Must War Find a way?” International Security 24 (2): 166–

198.Boesche, Roger. 2003. “Kautilya’s Arthasatra on War and Diplomacy in Ancient

India.” Journal of Military History 67: 9–38.Boland, Lawrence A. 1979. “Knowledge and the Role of Institutions in Economic

Theory.” Journal of Economic Issues 13 (4): 957–972.Booth, Ken. 2005. Critical Security Studies and World Politics. Boulder: Lynne

Rienner.Booth, Ken and Nicholas Wheeler. 2008. The Security Dilemma: Fear, Cooperation,

and Trust in World Politics. New York: Palgrave Macmillan.Boulding, Kenneth E. 1959. “National Images and International Systems.” Journal

of Conflict Resolution 3 (2): 120–131.Brodie, Bernard. 1959. Strategy in the Missile Age. Princeton: Princeton University

Press.

228 / bibliography

Brodie, Bernard. 1973. War and Politics. New York: Macmillan.Brooks, Stephen G. 1997. “Dueling Realisms.” International Organization 51 (3):

445–477.Brzezinski, Zbigniew. 1997. The Grand Chessboard: American Primacy and its

Geostrategic Imperatives. New York: Basic Books.Butterfield, Herbert. 1951. History and Human Relations. London: Collins.Buzan, Barry. 1991. People, States, and Fear. Boulder: Lynne Rienner.Byman, Daniel L. and Kenneth M. Pollack. 2001. “Let Us Now Praise Great Men:

Bringing the Statesman Back In.” International Security 25 (4): 107–146.Carr, E. H. 1940. The Twenty Years’ Crisis 1919–1939. London: McMillanChace, J. and C. Carr. 1988. America Invulnerable: The Quest for Absolute Security

from 1812 to Star Wars. New York: Summit Books.Chan, Steve. 2004. “Can’t Get no Satisfaction? The Recognition of Revisionist

State.” International Relations of the Asia-Pacific 4 (2): 207–238.Christensen, Thomas J. 1996. Useful Adversaries: Grand Strategy, Domestic

Mobilization, and Sino-American Conflict, 1947–1958. Princeton: Princeton University Press.

———. 1999. “China, the U.S.-Japan Alliance, and the Security Dilemma in East Asia.” International Security 23 (4): 49–80.

———. 2001. “Posing Problems Without Catching Up: China’s Rise and Challenges for U.S. Security Policy.” International Security 25 (4): 5–40.

———. 2002. “The Contemporary Security Dilemma: Deterring a Taiwan Conflict.” Washington Quarterly 25 (4): 7–21.

Colaresi, Michael. 2004. “When Doves Cry: International Rivalry, Unreciprocated Cooperation, and Leadership Turnover.” American Journal of Political Science 48 (3): 557–570.

Collins, Alan R. 1998. “GRIT, Gorbachev, and the End of the Cold War.” Review of International Studies 24 (2): 201–219.

———. 2000. The Security Dilemma of Southeast Asia. London: Macmillan.———. 2004. “State-induced Security Dilemma: Maintaining the Tragedy.”

Cooperation and Conflict 39 (1): 27–44.Cook, Karen S., Toshio Yamagishi, Coye Chershire, Robin Cooper, Masafumi

Matsuda, and Rie Mashima, 2005. “Trust Building via Risk Taking: A Cross-Societal Experiment.” Social Psychology Quarterly 68 (2): 121–142.

Copeland, Dale. 2000a. The Origins of Major War. Ithaca: Cornell University Press.

———. 2000b. “The Constructivism Challenge against Structural Realism: A Review Essay.” International Security 25 (2): 187–212.

———. 2001. “Theory and History in the Study of Major War.” Security Studies 10 (4): 212–239.

———. 2003. “A Realist Critique of the English School.” Review of International Studies 29: 427–441.

Cox, Michael. 2004. “Empire, Imperialism, and the Bush Doctrine.” Review of International Studies 30: 585–608.

Craig, Campbell. 2003. Glimmer of a New Leviathan: Total War in the Realism of Niebuhr, Morgenthau, and Waltz. New York: Columbia University Press.

Craig, Gordon A. and Alexander L. George. 1995. Force and Statecraft: Diplomatic Problems of Our Time. New York: Oxford University Press.

bibliography / 229

D’Lugo, David and Ronald Rogowski. 1993. “The Anglo-German Naval Race and Comparative Constitutional ‘Fitness’.” In The Domestic Bases of Grand Strategy, edited by Richard Rosecrance and Arthur A. Stein, 65–95. Ithaca: Cornell University Press.

Dasgupta, Partha. 1988. “Trust as a Commodity.” In Trust: Making and Breaking Cooperative Relations, edited by Diego Gambetta, 49–72. New York: Blackwell.

Desch, Michael C. 2001. “Liberals, Neocons, and Realcons.” Orbis 45 (4): 519–533.

———. 2003. “It is Kind to be Cruel: The Humanity of American Realism.” Review of International Studies 29: 415–426.

Desjardins, Marie-France. 1996. Rethinking Confidence-Building Measures. Adelphi Paper 307. London: International Institute of Strategic Studies.

Deutsch, Morton. 1958. “Trust and Suspicion.” Journal of Conflict Resolution 2 (4): 265–279.

Diamond, Jared. 1997. Guns, Germs, and Steel. New York: Norton.Dittmer, Lowell. 2004. “The Emerging Northeast Asian Regional Order.” In The

International Relations of Northeast Asia, edited by Samuel S. Kim, 331–362. Lanham: Rowman and Littlefield.

Doyle, Michael W. 1983. “Kant, Liberal Legacies, and Foreign Affairs.” Philosophy and Public Affairs, 12 (3 and 4): 205–235, 323–353.

Duffield, John. 2007. “What Are International Institutions?” International Studies Review 9: 1–22.

Duntley, Joshua D. 2005. “Adaptations to Dangers from Human.” In Handbook of Evolutionary Psychology, edited by David M. Buss. Hoboken: John Wiley & Sons.

Edelstein, David M. 2002. “Managing Uncertainty: Beliefs about Intentions and the Rise of Great Powers.” Security Studies 12 (1): 1–40.

Elman, Colin. 1996. “Horses for Courses: Why Not Neorealist Theory of Foreign Policy?” Security Studies 6 (1): 3–53.

———. 1995. “History vs. Neo-realism: A Second Look.” International Security 1: 182–195.

———. 1997. “Lakatose and Neorealism: A Reply to Vasquez.” American Political Science Review 91 (4): 923–926.

Etzioni, Amitai. 1962. The Hard Way to Peace: A New Strategy. New York: Crowell-Collier.

Evangelista, Matthew. 1993. “Internal and External Constraints on Grand Strategy: The Soviet Case.” In The Domestic Bases of Grand Strategy, edited by Richard Rosecrance and Author Stein, 154–178. Ithaca: Cornell University Press.

Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.” American Political Science Review 88 (3): 577–592.

———. 1995. “Rationalist Explanations for War.” International Organization 49 (3): 379–414.

Fearon, James D. and David Laitin. 1996. “Explaining Interethnic Cooperation.” American Political Science Review 90 (4): 715–735.

Feaver; Peter D., Gunther Hellman, Randall L. Schweller, Jeffery W. Taliaferro, William C. Wohlforth, Jeffery W. Lergo, and Andrew Moravcsik. 2000.

230 / bibliography

“Brother Can You Spare a Paradigm?” (Or Was Anybody Ever a Realist?) International Security 25 (1): 165–193.

Feldman, Lily Garner. 1993. “The Principle and Practice of ‘Reconciliation’ in German Foreign Policy: Relations with France, Israel, Poland, and the Czech Republic.” International Affairs 75 (2): 333–356.

Finel, Bernard I. 2001–2002. “Black Box or Pandora’s Box: Domestic Politics in Realism.” Security Studies 11 (2): 187–227.

Fischer, Markus. 1995–1996. “Machiavelli’s Theory of Foreign Politics.” Security Studies 5 (2): 248–279.

Fisher, Fritz. 1967. Germany’s Aims in the First World War. New York: Norton.———. 1974. World Power or Decline. New York: Norton.Forde, Steven. 1992. “Varieties of Realism: Thucydides and Machiavelli.” The

Journal of Politics 2: 372–393.———. 1995. “International Realism and the Science of Politics: Thucydides,

Machiavelli, and Neorealism.” International Studies Quarterly 2: 141–160.Frankel, Benjamin. 1996. “Restating the Realist Case: An Introduction.” Security

Studies 5 (3): ix–xx.Fravel, M. Taylor. 2005. “Regime Insecurity and International Cooperation:

Explaining China’s Compromises in Territorial Disputes.” International Security 30 (2): 46–83.

———. 2007. “Securing Borders: China’s Doctrine and Force Structure for Frontier Defense.” Journal of Strategic Studies 30 (4/5): 705–737.

———. 2008. Strong Borders, Secure Nation: Cooperation and Conflict in China’s Territorial Disputes. Princeton: Princeton University Press.

Gaddis, John Lewis. 1981. “Containment: Its Past and Future.” International Security 5 (4): 74–102.

———. 2005. The Strategy of Containment. Oxford: Oxford University Press.Gambetta, Diego. 1988. “Can We Trust Trust?” In Trust: Making and Breaking

Cooperative Relations, edited by Diego Gambetta, 213–237. New York: Blackwell.

George, Alexander L. 1969. “The ‘Operational Code’: A Neglected Approach to the Study of Political Leaders and Decision-Making.” International Studies Quarterly 13 (2): 190–222.

Giffin, Kim. 1967. “The Contribution of Studies of Source Credibility to a Theory of Interpersonal Trust in the Communication Process.” Psychological Bulletin 68 (2): 104–120.

Gilpin, Robert. 1981. War and Changes in World Politics. Cambridge: Cambridge University Press.

———. 1996. “No One loves a Political Realist.” Security Studies 5 (3): 3–26.Glaser, Charles L. 1992. “Political Consequences of Military Strategy: Expanding

and Refining the Spiral and Deterrence Models.” World Politics 44 (2): 497–538.———. 1994–1995. “Realists as Optimists: Cooperation as Self-help.”

International Security 19 (3): 50–90.———. 1997. “The Security Dilemma Revisted.” World Politics 50 (1): 171–201.———. 2003. “Structural Realism in a more Complex World.” Review of

International Studies 29: 403–414.———. 2004. “When Are Arms Races Dangerous.” International Security 28 (4):

44–84.

bibliography / 231

Glosny, Michael A. 2006. “Heading toward a Win-Win Future? Recent Developments in China’s Policy toward Southeast Asia.” Asian Security 2 (1): 24–57.

Goldgeier, James M. and Michael McFaul. 1992. “A Tale of Two Worlds: Core and Periphery in the Post-Cold War Era.” International Organization 46/2: 467–191.

Goldstein, Joshua S. and John R. Freeman. 1990. Three-Way Street: Strategic Reciprocity in World Politics. Chicago: University of Chicago Press.

Gray, Colin. 1993. Weapons Don’t Make War: Policy, Strategy, and Military Technology. Lawrence: University of Kansas Press.

Grieco, Joseph. 1988. “Anarchy and the Limits of Cooperation: A Realist Critique of the Newest Liberal Institutionalism.” International Organization 42 (3): 485–507.

———. 1990. Cooperation among Nations: Europe, America, and Non-Tariff Barriers to Trade. Ithaca: Cornell University Press.

———. 1993. “Understanding the Problem of International Cooperation: The Limits of Neoliberal Institutionalism and the Future of Realist Theory.” In Neorealism and Neoliberalism: The Contemporary Debate, edited by David Baldwin, 301–338. New York: Columbia University Press.

———. 1996. “State Interests and Institutional Rule Trajectories: A Neorealist Interpretation of the Maastricht Treaty and European Economic and Monetary Union.” Security Studies 5 (3): 261–305.

Grieco, Joseph, Robert Powell, and Duncan Snidal. 1993. “The Relative-Gains Problem for International Cooperation.” American Political Science Review 87 (3): 727–743.

Gu, Xiaosong and Brantly Womack. 2000. “Border Cooperation between China and Vietnam in the 1990s.” Asian Survey 40 (6): 1042–1058.

Haas, Ernst B. 1953. “The Balance of Power: Prescription, Concept, or Propaganda.” World Politics 5 (4): 442–477.

Haas, Mark L. 2005. The Ideological Origins of Great Power Politics, 1789–1989. Ithaca: Cornell University Press.

Hager, Robert P. Jr. and David A. Lake. 2000. “Balancing Empires: Competitive Decolonization in International Politics.” Security Studies 9 (3): 108–148.

Hardin, Russell. 1992. “The Street-level Epistemology of Trust.” Politics & Society 21 (4): 505–529.

———. 1996. “Trustworthiness.” Ethics 107 (1): 26–42.———. 2002. Trust and Trustworthiness. New York: Russell Sage.———. 2006. Trust. Cambridge: Polity.He, Kai. 2008. Institutional Balancing in the Asia Pacific: Economic Interdependence

and China’s Rise. London: Routledge.Hendrickson, David C. 2002. “Toward Universal Empire: The Dangerous Search

for Absolute Security.” World Policy Journal 19 (3): 1–10.Herz, John H. 1950. “Idealist Internationalism and the Security Dilemma.” World

Politics 2 (2): 157–180.———. 1951. Political Realism and Political Idealism: A Study in Theories and

Realities. Chicago: University of Chicago Press.———. 1961. International Politics in the Atomic Age. New York: Columbia

University Press.

232 / bibliography

Hobbes, Thomas. 1982 [1651]. Leviathan. Harmondsworth: Penguin.Hoddie, Matthew and Caroline Hartzell. 2005. “Signals of Reconciliation:

Institution-building and the Resolution of Civil Wars.” International Studies Review 7 (1): 21–40.

Hoffman, Aaron M., 2007. “The Structural Causes of Trusting Relationships: Why Rivals Do Not Overcome Suspicion Step by Step.” Political Science Quarterly 122 (2): 287–312.

Hui, Victoria Tin-bor. 2005. War and State Formation in Ancient China and Early Modern Europe. Cambridge: Cambridge University Press.

Huth, Paul. 1997. “Reputation and Deterrence: A Theoretical and Empirical Assessment.” Security Studies 7 (1): 72–99.

Ikenberry, G. John. 1998–1999. “Institutions, Strategic Restraint, and the Persistence of American Postwar Order.” International Security 22 (3): 43–78.

———. 2000. After Victory: Institutions, Strategic Restraint, and the Rebuilding of Order after Major Wars. Princeton: Princeton University Press.

———. 2002. “American Imperial Ambition.” Foreign Affairs 81 (5): 44–60.Jahn, Beate. 2005. “Kant, Mill, and Illiberal Legacies in International Affairs.”

International Organization 59: 277–207.Janis, Irving L. 1982. Groupthink: Psychological Studies of Policy Decisions and

Fiascoes. Boston: Houghton Mifflin.Jervis, Robert. 1970. The Logic of Images in International Relations. Princeton:

Princeton University Press.———. 1976. Perception and Misperception in International Politics. Princeton:

Princeton University Press.———. 1978. “Cooperation under the Security Dilemma.” World Politics 30 (2):

167–214.———. 1982. “Security Regimes.” International Organization 36 (2): 357–378.———. 1988. “Realism, Game Theory, and Cooperation.” World Politics 40 (4):

317–349.———. 1989a. “Rational Deterrence: Theory and Evidence.” World Politics 41:

183–207.———. 1989b. The Meaning of the Nuclear Revolution. Ithaca: Cornell University

Press.———. 1992. “The Political Implications of Loss Aversion.” Political Psychology

13 (1): 187–204.———. 1997. System Effects: Complexity in Political and Social Life. Princeton:

Princeton University Press.———. 1999. “Realism, Neoliberalism, and Cooperation: Understanding the

Debate.” International Security 24 (1): 42–63.———. 2001. “Was the Cold War a Security Dilemma?” Journal of Cold War

Studies 3 (1): 36–60.———. 2003. “Understanding the Bush Doctrine.” Political Science Quarterly

118 (3): 365–388.———. 2005. American Foreign Policy in a New Era. New York: Routledge.———. 2006. “The Remaking of a Unipolar World.” Washington Quarterly

29 (3): 7–19.Johnson, Chalmers. 2004. The Sorrows of Empire: Militarism, Secrecy, and the End

of the Republic. New York: Metropolitan Books.

bibliography / 233

Johnston, Alastair Iain. 1999. “Realism(s) and Chinese Security Policy in the Post-Cold War Period.” In Unipolar Politics: Relaism and State Strategies after the Cold War, edited by Ethan B. Kapstein and Michael Mastanduno, 261–318. New York: Columbia University Press.

Johnston, Alastair Iain and Robert S. Ross (eds.). 1999. Engaging China: The Management of an Emerging Power. London: Routledge.

Kaufman, Stuart J. 1996a. “An ‘International’ Theory of Inter-ethnic War.” Review of International Studies 22: 149–171.

———. 1996b. “Spiralling to Ethnic War: Elites, Masses, and Moscow in Moldova’s Civil War.” International Security 21 (2): 108–138.

———. 2001. Modern Hatred: the Symbolic Politics of Ethnic War. Ithaca, NY: Cornell University Press.

Kaufmann, Chaim D. 1996. “Possible and Impossible Solutions to Ethnic Civil Wars.” International Security 20 (4): 139–175.

Kautilya. 1969. Arthasatra. 2nd. ed., Translated and edited by R.P. Kangle. Bombay: University of Bombay Press.

Keohane, Robert O. 1984. After Hegemony: Cooperation and Discord in the World Political Economy. Princeton: Princeton University Press.

———. 1986a. “Theory of World Politics: Structural Realism and Beyond.” In Neorealism and Its Critics, edited by Robert O. Keohane, 158–203. New York: Columbia University Press.

———. 1986b. “Reciprocity in International Relations.” International Organization 40 (1): 1–27.

———. 1989. International Institutions and State Power: Essays in International Relations Theory. Boulder: Westview.

———. 1993. “Institutional Theory and the Realist Challenge after the Cold War.” In Neorealism and Neoliberalism: The Contempotary Debate, edited by David A. Baldwin, 269–300. New York: Columbia Univeristy Press.

Keohane, Robert O. and Lisa L. Martin. 1995. “The Promise of Institutionalist Theory.” International Security 20 (1): 39–51.

———. 2003. “Institutional Theory as a Research Program.” In Progress in International Relations Theory: Appraising the Field, edited by Colin Elman and Miriam F. Elman, 71–107. Cambridge, MA: MIT Press.

Keohane, Robert O. and Joseph S. Nye. 1977. Power and Interdependence. Boston: Little and Brown.

Khong, Yuen Foong. 1992. Analogies at War. Princeton: Princeton University Press.

Kier, Elizabeth A. 1997. Imagining War: French and British Military Doctrines between the Wars. Princeton: Princeton University Press.

Krasner, Stephen D. 1982a. “Structural Causes and Regimes Consequences: Regime as Intervening Variables.” International Organization 36 (2): 185–205

Krasner, Stephen D. 1982b. “Regimes and the Limits of Realism: Regimes as Autonomous Variables.” International Organization 36 (2): 497–510.

Kuhn, Thomas. 1970. The Structure of Scientific Revolution. Chicago: University of Chicago Press.

Kydd, Andrew. 1997a. “Game Theory and the Spiral Model.” World Politics 49 (2): 371–400.

234 / bibliography

Kydd, Andrew. 1997b. “Sheep in Sheep’s Clothing: Why Security Seekers Do Not Fight Each Other.” Security Studies 7 (1): 114–155.

———. 2000a. “Trust, Reassurance, and Cooperation.” International Organization 54 (2): 325–357.

———. 2000b. “Arms Races and Arms Controls: Modeling the Hawk Perspective.” American Journal of Political Science 44 (2): 228–244.

———. 2000c. “Overcoming Mistrust.” Rationality and Society 12 (4): 397–424.

———. 2005. Trust and Mistrust in International Relations. Princeton: Princeton University Press.

Labs, Eric J. 1997. “Beyond Victory: Offensive Realism and the Expansion of War Aims.” Security Studies 6 (4): 1–49.

Lakatos, Imre. 1970. “Falsification and the Methodology of Scientific Research Programmes.” In Criticism and the Growth of Knowledge: Proceedings of the International Colloquium in the Philosophy of Science, edited by Imre Lakatos and Alan Musgrave, 91–196. Cambridge: Cambridge University Press.

Lake, David A. 1992. Powerful Pacifists: Democratic States and War. American Political Science Review 86 (1): 24–37.

Lake, David A. and Robert Powell (eds.). 1999. Strategic Choice and International Relations. Princeton: Princeton University Press.

Larson, Deborah W. 1997. Anatomy of Mistrust: US-Soviet Relations during the Cold War. Ithaca: Cornell University Press.

Layne, Christopher. 2002–2003. “The Poster Child for Offensive Realism: America as a Global Hegemon.” Security Studies 12 (2): 120–164.

———. 2006a. “The Unipolar Illusion Revisited: The Coming End of the United States’ ‘Unipolar Moment.’ ” International Security 31 (2): 7–41.

———. 2006b. The Peace of Illusions: American Grand Strategy from 1940 to the Present. Ithaca: Cornell University Press.

Lebow, Richard Ned. 1981. Between Peace and War. Baltimore: John Hopkins University Press.

———. 1984. “Windows of Opportunity: Do States Jump Through Them?” International Security 9 (1): 147–186.

———. 1994. “The Long Peace, the End of the Cold War, and the Failure of Realism.” International Organization 48 (2): 249–277.

———. 1997. “Transitions and Transformations: Building International Cooperation.” Security Studies 6 (3): 154–179.

———. 2003. The Tragic Vision of Politics: Ethics, Interests, and Orders. Cambridge: Cambridge University Press.

———. 2008. A Cultural Theory of International Relations. Cambridge: Cambridge University Press.

Lee, Gerald Geunwook. 2002–2003. “To be Long or not To be Long—That is the Question: The Contradictions of Time-Horizon in Offensive Realism.” Security Studies 12 (2): 196–217.

Leffler, Melvin P. 1992. A Preponderance of Power: National Security, the Truman Administration and the Cold War. Stanford: Stanford University Press.

Legro, Jeffrey W. and Andrew Morvascik. 1999. “Is Anybody Still a Realist?” International Security 24 (2): 5–55.

bibliography / 235

Lewis, J. David and Andrew Weigert. 1985. “Trust as a Social Reality.” Social Forces 63 (4): 967–985.

Levy, Jack S. 1984. “The Offensive/Defensive Balance of Military Technology: A Theoretical and Historical Analysis.” International Studies Quarterly 28 (2): 219–238.

———. 1987. “Declining Power and the Preventive Motivation for War.” World Poliics 40 (1): 82–107.

———. 1990–1901. “Preferences, Constraints, and Choices in July 1914.” International Security 15 (3): 151–186.

———. 1997. “Prospect Theory, Rational Choice, and International Relations.” International Studies Quarterly 41 (1): 87–112.

———.2003. “Balances and Balancing: Concepts, Propositions, and Research Design,” in John A. Vasquez and Collin Elman eds., Realism and the Balance of Power: A New Debate (Upper Saddle River, N. J.: Printice Hall), pp. 228–147.

Levy, Jack S. and William R. Thompson. 2005. “Hegemonic Threat and Great Power Balancing in Europe, 1495–1999.” Security Studies 15 (1): 1–33.

Lieber, Kier A. 2005. War and Engineers: The Primacy of Politics over Technology. Ithaca, NY: Cornell University Press.

Lieber, Kier A. 2007. “The New History of World War I and What It Means for International Relations Theory.” International Security 32 (2): 155–191.

Lindskold, Svenn. 1978. “Trust Development, the GRIT Proposal, and the Effects of Conciliatory Acts on Conflict and Cooperation.” Psychological Bulletin 85 (4): 772–793.

Lipson, Charles. 1984. “International Cooperation in Economic and Security Affairs.” World Politics 37 (1): 1–23.

Liu, Feng, and Ruizhuang Zhang. 2006. “The Typologies of Realism.” Chinese Journal of International Politics 1 (1): 109–134.

Lobell, Steven E. 2002–2003. “War Is Politics: Offensive Realism, Domestic Politics, and Security Strategy.” Security Studies 12 (2): 165–195.

Lobell, Steven E., Norrin M. Ripsman, and Jeffrey W. Taliaferro (eds.). 2009. Neoclassical Realism, the State, and Foreign Policy. Cambridge: Cambridge University Press.

Lynn-Jones, Sean M. 1998. “Realism and America’s Rise.” International Security 23 (2): 157–182.

MacDonald, Paul K. 2003. “Useful Fiction or Miracle Maker: The Competing Epistemological Foundations of Rational Choice.” Theory American Political Science Review 97 (4): 551–565.

Machiavelli, Niccolò, 1997 [1532]. The Prince, translated and edited by Angelo M. Codevilla. New Haven: Yale University Press.

Mansfield, Edward D. and Jack Snyder. 1005. Electing to War: Why Emerging Democracies Go to War. Cambridge, MA: MIT Press.

Maoz, Zeev and Dan S. Felsenthal. 1987. “Self-Binding Commitments, the Inducement of Trust, Social Choice, and the Theory of International Cooperation.” International Studies Quarterly 31 (2): 177–200.

Markey, Daniel. 1999. “Prestige and the Origins of War.” Security Studies 8 (4): 126–173.

236 / bibliography

Mastanduno, Michael. 1997. “Preserving the Unipolar Moment: Realist Theories and U.S. Grand Strategy after the Cold War.” International Security 21 (4): 49–88

Maxwell, Neville. 1970. India’s China War. New York: Pantheon Books.McMahon, Robert. 1991. “Credibility and World Power: Exploring the

Psychological Dimension in Postwar American Diplomacy.” Diplomatic History 15 (4): 455–471.

Mearsheimer, John J. 1990. “Back to the Future: Instability in Europe after the Cold War.” International Security 15 (1): 5–56.

———. 1994–1995. “The False Promise of International Institutions.” International Security 19 (3): 5–49.

———. 2001. The Tragedy of Great Power Politics. New York: Norton.———. 2006. “Interview.” International Relations 20 (1): 105–123.Mearsheimer, John J. and Stephen Walt. 2003. “An Unnecessary War.” Foreign

Policy (January–February): 51–59.Medeiros Evan. 2005–2006. “Strategic Hedging and the Future of Asia-Pacific

Stability.” Washington Quarterly 29 (1): 145–168.Men, Honghua. 2005. Baquan zhiyi: mwi guo guoji zhidu zhanlue [Wings of

Hegemony: America’s Strategy on International Institutions]. Beijing: Peking University Press.

Mercer, Jonathan. 1996. Reputation in International Politics. Ithaca: Cornell University Press.

Midford, Paul. 2002. “The Logic of Reassurance and Japan’s Grand Strategy.” Security Studies 11 (3): 1–43.

Miller, Benjamin. 1995. When Opponents Cooperate: Great Power Conflict and Collaboration in World Politics. Ann Arbor: University of Michigan Press.

———. 2000. “Explaining Variations in Regional Peace: Three Strategies for Peace-making.” Cooperation and Conflict 35 (2): 155–191.

———. 2001. “The Global Sources of Regional Transitions from War to Peace.” Journal of Peace Research 38 (2): 199–225.

Milner, Helen. 1992. “International Theories of Cooperation among Nations: Strengths and Weaknesses.” World Politics 44 (3): 466–496.

Mitchell, Christopher R. 1991. “A Willingness to Talk: Conciliatory Gestures and De-Escalation.” Negotiation Journal 7 (4): 405–430.

Mitzen, Jennifer. 2006. “Ontological Security in World Politics: State Identity and Security Dilemma.” European Journal of International Relations 12 (3): 341–370.

Modelski, George, 1987. Long Cycles in World Politics. Seattle: University of Washington Press.

Montgomery, Evan Braden. 2006. “Breaking out of the Security Dilemma: Realism, Reassurance, and the Problem of Uncertainty.” International Security 31 (2): 151–185.

———. 2008. “Risking Peace to Prepare for War: Mobilization, Statecraft, and the Logic of Counterfeit Diplomacy.” Paper presented at the 49th Annual Convention of International Studies Association, San Francisco, March 26–30.

Moravcsik, Andrew. 2003. “Liberal International Relations Theory: A Scientific Assessment.” In Progress in International Relations Theory: Appraising the Field, edited by Colin Elman and Miriam F. Elman, 159–204. Cambridge, MA: MIT Press.

Morgenthau, Hans J. 1946. Scientific Man versus Power Politics. Chicago: University of Chicago Press.

bibliography / 237

———. 1948. Politics among Nations. New York: Knopf.———. 1967. Politics among Nations (4th ed.). New York: Knopf.Nelson, Anna Katsen. 2005. “Continuity and Change in the Age of Unlimited

Power.” Diplomatic History 29 (3): 437–439.Niebuhr, Reinhold. 1960 [1932]. Moral Man and Immoral Society: A Study in

Ethics and Politics. New York: Charles Scribner’s Sons.Offner, Arnold A. 2005. “Rogue Nation, Rogue State: Bush and U.S. National

Security.” Diplomatic History 29 (3): 433–435.Oneal, John R. and Bruce M. Russett 1999. “The Kantian Peace: The Pacific

Benefits of Democracy, Interdependence, and International Organizations, 1885–1992.” World Politics 52 (1): 1–37.

Organski, A. F. L. and Jack Kugler. 1980. The War Ledger. Chicago: University of Chicago Press.

Osgood, Charles A. 1962. An Alternative to War or Surrender. Urbana: University of Illinois Press.

Oye, Kenneth A. (ed.). 1986. Cooperation under Anarchy. Princeton: Princeton University Press.

Palan, Ronen. 2000. “A World of Their Making: An Evaluation of the Constructivist Critique in International Relations.” Review of International Studies 26 (4): 575–598.

Peterson, Sue, Michael J. Tierney, and Dan Maliniak. 2005. “Teaching and Research Practices, Views on the Discipline, and Policy Attitudes of IR Faculty at U.S. Colleges and Universities.” Manuscript, Reves Center for International Studies, William Mary College.

Porch, Douglas. 2000. “Military ‘Culture’ and the Fall of France in 1940: A Review Essay.” International Security 24 (4): 157–180.

Posen, Barry. 1984. The Sources of Military Doctrines: France, Britain, and Germany between the World Wars. Ithaca: Cornell University Press.

———. 1993. “The Security Dilemma and Ethnic Conflict.” Survival 35 (1): 27–47.Powell, Robert. 1991. “Absolute and Relative Gains in International Relations

Theory.” American Political Science Review 85: 1303–1320.———. 1994. “Anarchy in International Relations Theory: The Neorealist-

Neoliberal Debate.” International Organization 48 (2): 313–344.———. 2006. “War as a Commitment Problem.” International Organization

60 (2): 169–203.Press, Daryl. 2005. Calculating Credibility: How Leaders Assess Military Threat

during Crisis. Ithaca: Cornell University Press.Prestowitz, Clyde. 2003. Rogue Nation: American Unilateralism and the Failure of

Good Intentions. New York: Basic Books.Price, Russell. 1982. “Ambizione in Machiavelli’s Thought.” History of Political

Thought 3 (3): 383–445.Rathbun, Brian C. 2007. “Uncertain about Uncertainty: Understanding the

Multiple Meanings of a Crucial Concept in International Relations Theory.” International Studies Quarterly 51: 533–557.

Reiter, Dan. 1995. “Exploring the Power Keg Myth: Preemptive Wars Almost Never Happen.” International Security 20 (2): 5–34.

Rendall, Matthew. 2006. “Defensive Realism and the Concert of Europe.” Review of International Studies 32: 523–540.

238 / bibliography

Resnick, Evan. 2001. “Defining Engagement.” Journal of International Affairs 54 (2): 551–566.

Ripsman, Norrin M. 2005. “Two Stages of Transition from a Region of War to a Region of Peace: Realist Transition and Liberal Endurance.” International Studies Quarterly 49 (4): 669–693.

Ripsman, Norrin M., and Jack S. Levy. 2007. “The Preventive War that Never Happened: Britain, France and the Rise of Germany in the 1930s.” Security Studies 16 (1): 32–67.

Roe, Paul. 2000. “Former Yugoslavia: The Security Dilemma That Never Was.” European Journal of International Relations 6 (3): 373–393.

———. 2001. “Actors’ Responsibility in Tight, Regular, or Loose Security Dilemmas.” Security Dialogue 32 (1): 103–116.

———. 2004. “Which Security Dilemma? Mitigating Ethnic Conflict: The Case of Croatia.” Security Studies 13 (4): 280–313.

———. 2005. Ethnic Violence and the Societal Security Dilemma. London: Routledge.

Rose, Gideon. 1998. “Neoclassical Realism and Theories of Foreign Policy.” World Politics 51 (1): 144–172.

Rosecrance, Richard. 2001. “Has Realism Become Cost-Benefit Analysis? A Review Essay.” International Security 26 (2): 132–154.

———. 2002. “War and Peace.” World Politics 52 (1): 137–166.Rosecrance, Richard and Arthur A. Stein. 1993. “Beyond Realism: The Study of

Grand Strategy.” In The Domestic Bases of Grand Strategy, edited by Richard Rosecrance and Arthur A. Stein, 3–21. Ithaca: Cornell University Press.

Ross, Lee and Andrew Ward. 1995. “Psychological Barriers to Dispute Resolutions.” In Advances in Experimental Social Psychology Vol. 27, edited by M.P. Zanna, 255–304. San Diego, CA: Academic Press.

Ross, Robert S. 1995. Negotiating Cooperation: The United States and China, 1969–1989. Stanford: Stanford University Press.

Ruggie, John G. 1983. “Continuity and Transformation in the World Polity: Toward a Neorealist Synthesis.” World Politics 25 (2): 261–285.

Samuels, Richard J. 2003. Machiavelli’s Children: Leaders and Their Legacies in Italy and Japan. Ithaca: Cornell University Press.

Schelling, Thomas. 1966. Arms and Influence. New Haven: Yale University Press.Schroeder, Paul W. 2005. “Embedded Counterfactuals and World War I as an

Unavoidable War.” Unpublished manuscript, Ohio State University.Schultz, Kenneth. 2005. “The Politics of Risking Peace: Do Hawks or Doves

Deliver the Olive Branch?” International Organization 59 (1): 1–38.Schweller, Randall L. 1992. “Domestic Structures and Preventive War: Are

Democracies more Pacific?” World Politics 44: 235–239.———. 1993. “Deadly Imbalance: Tripolarity and WWII.” International Studies

Quarterly 37 (1): 73–103.———. 1994. “Bandwagoning for Profit: Bring the Revisionist State back In.”

International Security 19 (1): 72–107.———. 1996. “Neorealism’s Status Quo Bias: What Security Dilemma?” Security

Studies 5 (3): 91–121.———. 1998. Deadly Imbalances: Tripolarity and Hitler’s Strategy of World

Conquest. New York: Columbia.

bibliography / 239

———. 1999. “Managing the Rise of Great Powers: History and Theory.” In Engaging China: The Management of an Emerging Power, edited by Alastair Iain Johnston and Robert S. Ross, 1–31. London: Routledge.

———. 2001. “The Problem of International Order Revisited: A Review Essay.” International Security 26 (1): 161–186.

———. 2003. “The Progressiveness of Neoclassical Realism.” In Progress in International Relations Theory, edited by Colin Elman and Miriam F. Elman, 311–347. Cambridge, MA: MIT Press.

———. Unanswered Threats: Political Constraints on the Balance of Power. Princeton: Princeton University Press.

———. 2009. “Neoclassical Realism and State Mobilization: Expansionist Ideology in the Age of Mass Politics.” In Neoclassical Realism, the State, and Foreign Policy, edited by Lobell, Steven E., Norrin M. Ripsman, and Jeffrey W. Taliaferro, 227–250. Cambridge: Cambridge University Press.

Schweller, Randall L. and David Priess. 1997. “A Tale of Two Realisms: Expanding the Institutions Debate.” Mershon International Studies Review 41: 1–32.

Schweller, Randall L. and William Wohlforth. 2000. “Power Test: Evaluating Realism in Response to the End of the Cold War.” Security Studies 9 (3): 60–107.

Shambaugh, David. 2004–2005. “China Engages Asia: Reshaping the Regional Order.” International Security 29 (3): 64–99.

Shang, Yang. ~339 B. C. Shang Jun Shu [Books of Lord Shang].Snyder, Glenn H. 1984. “The Security Dilemma in Alliance Politics.” World

Politics, 36 (4): 461–495.———. 1996. “Process Variables in Neorealist Theory.” Security Studies 5 (3):

167–192.———. 1997. Alliance Politics. Ithaca: Cornell University Press.———. 2002. “Mearsheimer’s World: Offensive Realism and the Struggle for

Security.” International Security 27 (1): 149–173.Snyder, Jack. 1984. “Civil-Military Relations and the Cult of the Offensive, 1914

and 1984.” International Security 9 (1): 108–146.———. 1985. “Perceptions of the Security Dilemma in 1914.” In Psychology and

Deterrence, edited by Robert Jervis, Richard Ned Lebow, and Janice Gross Stein, 153–179. Baltimore, John Hopkins University Press.

———. 1991. Myths of Empire. Ithaca: Cornell University Press.Snyder, Jack and Lieber, Kier A. 2008. “Correspondence: Defensive Realism and

the ‘New’ History of World War I.” International Security 33 (1): 174–185.Spirtas, Michael. 1996. “A House Divided: Tragedy and Evil in Realist Theory.”

Security Studies 5 (3): 385–423.Spykman, Nicholas John. 1942. America’s Strategy in World Politics: The United

States and the Balance of Power. New York: Harcourt, Brace.Stein, Arthur A. 1982. “Coordination and Collaboration: Regimes in an Anarchic

World.” International Organization 36 (2): 299–324.Stein, Janice Gross. 1988. “Building Politics into Psychology: The Misperception

of Threat.” Political Psychology 9 (2): 245–271.———. 1991. “Reassurance in International Conflict Management.” Political

Science Quarterly 106 (3): 431–451.Taliaferro, Jeffery W. 2000–2001. “Security Seeking under Anarchy: Defensive

Realism Revisited.” International Security 25 (3): 128–161.

240 / bibliography

Zhang, Yunling and Shiping Tang, 2005. “China’s Regional Strategy,” in David Shambaugh ed., Power Shift: China and Asia’s New Dynamics (Berkeley: University of California Press, 2005), pp. 48–68.

Taliaferro, Jeffery W. 2001. “Realism, Power Shifts, and Major War.” Security Studies 10 (4): 145–178.

———. 2004. Balancing Risks: Great Power Intervention in the Periphery. Ithaca: Cornell University Press.

Tang, Shiping. 2003. Construct China’s Ideal Security Environment (in Chinese). Beijing, China: China Social Sciences Press.

———. 2004. “A Systemic Theory of the Security Environment.” Journal of Strategic Studies 27 (1): 1–32.

———. 2005. “Reputation, Cult of Reputation, and International Conflict.” Security Studies 24 (4): 34–62.

———. 2007. “Correspondence: Uncertainty and Reassurance in International Politics.” International Security 27 (3): 180–183.

———. 2008a. “Fear in International Politics: Two Positions.” International Studies Review 10 (3): 451–470.

———. 2008b. “From Offensive to Defensive Realism: A Social Evolutionary Interpretation of China’s Security Strategy.” In China’s Ascent: Power, Security, and the Future of International Politics, edited by Robert Ross and Zhu Feng, 141–162. Ithaca: Cornell University Press.

———. 2008c. “The Social Evolutionary Psychology of Fear (and Trust): Or why international cooperation is difficult.” Paper presented at the 49th Annual Convention of International Studies Association in San Francisco, March 26–29.

———. 2009a. “The Security Dilemma: A Conceptual Analysis.” Security Studies 18 (3): 587–623.

———. 2009b. “Social Evolution of International Politics: From Mearsheimer to Jervis.” European Journal of International Relations (forthcoming).

———. 2009c. “Taking Stock of Neoclassical Realism.” International Studies Review 11 (4): 798–803.

———. 2009d. “Dimensions of Uncertainty and Their Cognitive Challenges: A Social Evolutionary Perspective.” Paper presented at the American Political Science Association Meeting, Toronto, Canada, September 3–6, 2009.

———. 2010a. “The Security Dilemma and Ethnic Conflict: Toward a Dynamic and Integrative Theory of Ethnic Conflict.” Review of International Studies (forthcoming).

———. 2010b. “Offense-Defense Theory: Toward a Definitive Understanding.” Chinese Journal of International Politics (forthcoming).

Thomas Hobbes, 1968 [1651], Leviathan, London: Penguin.Thucydides. 1972. The Peloponnesian War, trans. Rex Warner. London: Penguin.Toft, Monica Duffy. 2003. The Geography of Ethnic Violence: Identity, Interests,

and Territory. Princeton: Princeton University Press.Trachtenberg, Marc. 1990. “The Meaning of Mobilization in 1914.” International

Security 15 (3): 120–150.———. 1991. History and Strategy. Princeton: Princeton University Press.———. 2003. “The Question of Realism: A Historian’s View.” Security Studies

13 (1): 156–194.

bibliography / 241

Treisman, Daniel. 2004. “Rational Appeasement.” International Organization 58 (1): 345–373.

Van Evera, Stephen. 1984. “The Cult of the Offensive and the Origins of the First World War.” International Security 9 (1): 58–107.

———. 1985. “Why Cooperation Failed in 1914?” World Politics 38 (1): 80–117.———. 1990–1901. “Primed for Peace: Europe after the Cold War.” International

Security 15 (3): 7–57.———. 1994. “Hypotheses on Nationalism and War.” International Security

18 (4): 5–39.———. 1999. Causes of War: Power and the Roots of Conflict. Ithaca: Cornell

University Press.Wade, Robert. 1990. Governing the Market: Economic Theory and the Role of

Government in East Asian Industrialization. Princeton: Princeton University Press.

Wallander, Celeste A. and Robert O. Keohane. 1999. “Risk, Threat, and Security Institutions.” In Imperefect Union: Security Institutions over Time and Space, edited by Helga Haftendorn, Robert O. Keohane, and Celeste A. Wallander, 21–47. New York: Oxford Unviersity Press.

Walt, Stephen M. 1987. The Origins of Alliance. Ithaca: Cornell University Press.———. 1999. “Rigor or Rigor Mortis: Rational Choice and Security Studies.”

International Security 23 (4): 5–58.———. 2005. Taming American Power: The Global Response to U.S. Primacy. New

York: Norton.Walter, Barbara F. 2001. Committing to Peace: The Success Settlement of Civil Wars.

Princeton: Princeton University Press.Waltz, Kenneth N. 1959. Man, the State and War. New York: Columbia University

Press.———. 1967. “The Politics of Peace.” International Studies Quarterly 11 (3):

199–211.———. 1979. Theory of International Politics. Reading: Addision-Wesley.———. 1986. “Reflections on Theory of International Politics: A Respone to My

Critics.” In Neorealism and its Critics, edited by Robert O. Keohane, 322–346. New York: Columbia University Press.

———. 1988. “The Origins of War in Neorealist Theory.” Journal of Interdisciplinary History 18 (4): 615–628.

———. 1990. “Nuclear Myths and Political Realities.” American Political Science Review 84 (3): 731–745.

———. 1993. “The Emerging Structure of International Politics.” International Security 18 (2): 44–79.

———. 1996. “Theory of International Politics Is not Theory of Foreign Policy.” Security Studies 6 (1): 54–57.

———. 2000. “Structural Realism after the Cold War.” International Security 25 (1): 5–41.

Weinberger, Seth. 2003. “Institutional Signaling and the Origin of the Cold War.” Security Studies 12 (4): 80–115.

Welch, David A. 2003. “Why International Relations Theorists Should Stop Reading Thucydides.” Review of International Studies 29: 301–319.

242 / bibliography

Wendt, Alexander. 1987. “The Agent-Structure Problem in International Relations Theory.” International Organization 41 (3): 335–370.

Wendt, Alexander. 1992. “Anarchy Is What States Makes of It.” International Organization 46 (2): 391–425.

———. 1995. “Constructing International Politics.” International Security 20 (1): 71–81.

———. 1999. Social Theories of International Politics. Cambridge: Cambridge University Press.

Wheeler, Nicholas and Ken Booth. 1992. “The Security Dilemma.” In Dilemma of World Politics, edited by John Baylis and Nicholas Wheeler, 29–60. Oxford: Oxford University Press.

———. 2008. The Security Dilemma: Fear, Cooperation, and Trust in World Politics. Houndmills, Basingstoke: Palgrave Macmillan.

Whiting, Allen. 1975. The Chinese Calculus of Deterrence: India and Indochina. Ann Arbor: University of Michigan Press.

Wilson, E.O. 1975. Sociobiology: The New Synthesis. Cambridge, MA: Belknap Press of Harvard University Press.

Wohlforth, William C. 1994–1995. “Realism and the End of the Cold War.” International Security 19 (3): 91–129.

Wolfers, Arnold. 1951. “The Pole of Power and the Pole of Indifference.” World Politics 4 (1): 39–63.

———. 1952. “ ‘National Security’ as an Ambiguous Symbol.” Political Science Quarterly 67 (4): 481–502.

———. 1962. Discord and Collaboration: Essay on International Politics. Baltimore: John Hopkins University Press.

Womack, Brantly. 2006. China and Vietnam: The Politics of Asymmetry. Cambridge: Cambridge University Press.

Yamagishi, Toshiol; Kanazawa, Satoshi; Mashima, Rie; and Terai, Shigeru. 2005. “Separating Trust from Cooperation in a Dynamic Relationship: Prisoner’s Dilemma with Variable Dependence.” Rationality and Society 17 (3): 275–308.

Zakaria, Fareed. 1992. “Realism and Domestic Politics.” International Security 17 (1): 177–198.

———. 1998. From Wealth to Power: The Unusual Origins of America’s World Role. Princeton: Princeton University Press.

Ziring, Lawrence, Jack. C. Plano, and Roy Olton. 1995. International Relations: A Political Dictionary. Santa Barbara: ABC-CLIO.

Africa, 34, 186, 207Appeasement, 101, 103, 104, 106, 125,

134, 144, 171, 208, 213, 218, 220, 240

Balance/balancing, 12, 15, 16, 17, 18, 20, 21, 30, 71, 76, 100, 106, 109, 111, 118, 120, 136, 145, 146, 159, 166, 167, 169, 172, 182, 194, 196, 199, 206, 207, 213, 214, 215, 217, 220, 231, 235, 238, 239, 240

Balance-of-power, 106, 194, 214Bandwagoning, 100, 111, 155, 213,

214, 238Booth, Ken, 33, 47, 49, 50, 56, 57,

60, 61, 69, 91, 200, 205, 217, 227, 242

Boulding, Kenneth, 44, 50, 202, 203, 204, 227

Britain (Great Britain), 3, 25, 50, 63, 79, 93, 151, 154, 156, 167, 172, 185, 186, 190, 198, 212, 214, 237, 238

Butterfield, Herbert, v, 33, 34, 35, 36, 37, 38, 39, 56, 73, 74, 175, 200, 201, 202, 228

Carr, E. H., 24, 26, 173, 224, 228China, 2, 25, 34, 54, 67, 69, 77, 84,

85, 86, 134, 141, 147, 150, 151, 152, 153, 154, 157, 174, 186, 190, 193, 207, 208, 213, 214, 215, 216, 217, 224, 227, 228, 230, 231,

232, 233, 235, 238, 239, 240, 242

Cold War, the, vii, 2, 34, 66, 86, 89, 117, 131, 134, 144, 150, 181, 185, 186, 187, 202, 207, 208, 215, 226, 228, 232, 233, 235, 236, 239, 241, 242

Common interest, 20, 21, 112, 113, 134, 138, 165, 214

Confidence-building measures, 117, 122, 161, 229

Conflict of interest, ix, 43, 44, 45, 46, 47, 51, 62, 79, 106, 111, 114, 115, 116, 122, 134, 149, 150, 153, 186, 202, 203

Containment, 100, 102, 103, 104, 121,125, 172, 213, 216, 230

Cooperation, vii, 2, 6, 15, 16, 20, 22, 23, 30, 33, 45, 47, 63, 70, 89, 99, 101, 102, 103, 104, 107, 110, 112, 113, 114, 115, 116, 117, 118, 122, 123, 124, 126, 127, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 147, 149, 151, 152, 153, 154, 155, 156, 157, 159, 160, 161, 162, 165, 166, 167, 168, 170, 171, 172, 180, 181, 182, 195, 197, 198, 200, 202, 215, 217, 218, 219, 220, 221, 222, 224, 226, 227, 230, 231, 232, 233, 235, 236, 237, 238, 240, 241, 242

Copeland, Dale, 12, 16, 17, 18, 33, 34, 62, 78, 86, 89, 90, 100, 102, 106,

Index

244 / index

Copeland, Dale—Continued 108, 147, 148, 169, 170, 171, 194,

195, 199, 201, 204, 207, 208, 210, 211, 212, 215, 217, 223, 228

Credibility, 6, 13, 68, 135, 136, 137, 148, 150, 151, 218, 219, 221, 230, 236, 237

Defensive realist state, 1, 3, 6, 9, 15, 23, 24, 25, 26, 27, 28, 29, 31, 32, 39, 47, 48, 52, 53, 56, 57, 58, 61, 62, 63, 64, 65, 70, 71, 76, 78, 79, 80, 81, 82, 83, 85, 86, 87, 88, 89, 90, 91, 92, 94, 95, 96, 103, 107, 108, 110, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 127, 130, 142, 143, 147, 148, 156, 157, 158, 159, 179, 180, 182, 187, 198, 199, 200, 203, 207, 211, 213, 215, 216, 219, 220, 221, 225

Distrust, 132, 218, 219Domestic politics, 3, 4, 10, 15, 18, 23,

69, 107, 124, 131, 161, 171, 194, 196, 198, 208, 224, 230, 235, 242

Edelstein, David M., 101, 143, 144, 145, 218, 220, 229

Empire, 25, 28, 32, 86, 87, 93, 106, 109, 150, 173, 174, 186, 197, 198, 199, 208, 225, 227, 228, 231, 232, 239

Engagement, 100, 101, 102, 103, 104, 121, 124, 125, 133, 134, 142, 144, 213, 219, 220, 238

Etzioni, Amitai, 130, 229

Fear, 12, 18, 19, 32, 35, 36, 38, 39, 40, 41, 42, 48, 49, 50, 59, 64, 67, 69, 75, 77, 81, 82, 83, 86, 94, 108, 125, 130, 132, 144, 156, 157, 158, 162, 180, 183, 186, 197, 199, 201, 202, 203, 208, 209, 210, 218, 220, 223, 225, 227, 228, 240, 242

Fearon, James D., 70, 148, 166, 209, 212, 229

France, 17, 50, 63, 77, 79, 85, 90, 154, 155, 156, 167, 172, 185, 186, 197, 214, 229, 230, 237, 238\

Germany, 3, 27, 36, 50, 53, 68, 74, 77, 79, 85, 86, 87, 90, 93, 116, 144, 145, 154, 156, 167, 172, 175, 185, 186, 190, 197, 212, 214

Gilpin, Robert, 10, 21, 28, 92, 106, 169, 170, 194, 211, 223, 230

Glaser, Charles, 3, 9, 12, 18, 19, 20, 21, 23, 26, 27, 30, 33, 34, 41, 47, 48, 51, 52, 54, 58, 59, 60, 62, 63, 64, 71, 73, 75, 80, 81, 83, 88, 89, 90, 91, 94, 102, 103, 108, 110, 112, 113, 117, 120, 121, 122, 123, 125, 127, 130, 139, 140, 141, 143, 144, 145, 146, 148, 161, 166, 167, 170, 172, 195, 196, 198, 200, 204, 205, 207, 208, 209, 214, 218, 219, 221, 222, 223, 230

Gorbachev, 141, 187, 216, 226, 228

Han Fei Tzu, 173, 174, 224Hedging, 102, 104, 105, 125, 133,

145, 213, 236Herz, John, v, 33, 34, 35, 231Hitler, 32, 50, 80, 86, 90, 123, 134,

200, 208, 214Hobbes, Thomas, 232, 240Human nature, 9, 11, 13, 170, 195, 201

Imperialism, 18, 165, 170, 171, 190, 223, 228

Incompatibility (illusory or real, subjective or objective), 44, 45, 49, 50, 51, 60, 115, 116, 202, 203, 204, 205

India, 54, 77, 153, 155, 174, 190, 216, 227, 242

Institution, 33, 69, 108, 126, 130, 153, 166, 167, 168, 169, 170, 194, 212,

index / 245

215, 217, 219, 221, 223, 225, 231, 232, 233, 236, 239, 241

Intention, 6, 11, 14, 15, 19, 27, 31, 32, 35, 36, 37, 38, 39, 40, 42, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 56, 57, 58, 59, 60, 61, 63, 64, 65, 70, 71, 76, 77, 78, 79, 80, 81, 82, 83, 86, 87, 93, 102, 103, 107, 108, 110, 111, 113, 114, 115, 116, 119, 121, 122, 123, 124, 125, 126, 127, 129, 130, 131, 132, 134, 135, 137, 138, 139, 140, 142, 143, 144, 145, 147, 148, 149, 151, 152, 153, 155, 156, 158, 160, 161, 166, 171, 172, 175, 180, 183, 185, 187, 195, 200, 201, 202, 204, 205, 209, 210, 211, 212, 213, 215, 216, 217, 218, 219, 220, 221, 225, 229, 237

Jervis, Robert, v, 3, 13, 14, 15, 18, 19, 20, 21, 24, 25, 30, 33, 34, 35, 37, 38, 39, 40, 41, 42, 44, 47, 48, 49, 50, 54, 56, 58, 62, 63, 66, 67, 68, 69, 74, 75, 80, 84, 85, 86, 87, 88, 89, 93, 96, 103, 110, 112, 113, 117, 118, 124, 125, 130, 131, 136, 144, 145, 146, 148, 153, 155, 156, 162, 166, 167, 168, 170, 172, 175, 181, 192, 195, 196, 197, 198, 200, 201, 202, 204, 205, 206, 207, 209, 210, 211, 214, 215, 216, 217, 218, 219, 221, 222, 223, 224, 226, 232, 239, 240

Kautilya, 173, 174, 233Keohane, Robert O., 33, 134, 137,

152, 166, 168, 169, 170, 194, 197, 222, 223, 227, 233, 241

Khrushchev, 66, 69, 150, 153, 187, 221, 222, 226

Kydd, Andrew, xiv, 9, 14, 24, 33, 34, 40, 41, 53, 62, 66, 67, 69, 70, 73, 79, 80, 81, 82, 83, 85, 89, 92, 93,

113, 117, 118, 119, 120, 121, 123, 125, 129, 130, 133, 135, 136, 137, 139, 140, 141, 142, 143, 144, 147, 154, 156, 158, 161, 166, 167, 169, 170, 186, 187, 195, 196, 201, 202, 204, 205, 209, 210, 215, 216, 218, 219, 222, 224, 233, 234

Labs, Eric J., 14, 15, 17, 19, 21, 22, 26, 29, 106, 107, 108, 129, 170, 197, 204, 221, 224, 234

Larson, Deborah, 9, 13, 117, 130, 135, 136, 137, 139, 140, 141, 148, 150, 151, 153, 158, 161, 166, 210, 218, 219, 221, 222, 226, 234

Layne, Christopher, 4, 14, 16, 19, 89, 182, 196, 234

Lebow, Richard Ned, xiv, 86, 88, 89, 130, 195, 211, 224, 234, 239

Legro, Jeffrey, xiv, 3, 4, 9, 10, 19, 193, 196, 198, 213, 234

Lynn-Jones, Sean, 235

Machiavelli, 12, 108, 109, 173, 174, 224, 230, 235

Mearsheimer, John M., xiii, 2, 3, 4, 9, 12, 15, 16, 17, 18, 19, 20, 21, 22, 24, 26, 28, 29, 30, 33, 58, 87, 88, 89, 99, 101, 105, 106, 107, 108, 109, 112, 124, 129, 130, 156, 169, 170, 173, 175, 181, 190, 194, 195, 196, 199, 204, 208, 211, 213, 214, 215, 223, 224, 226, 236, 240

Mercer, Jonathan, xiv, 13, 68, 136, 162, 208, 236

Moravcsik, Andrew, 3, 4, 9, 10, 19, 193, 196, 198, 213, 229, 236

Morgenthau, Hans J., 12, 24, 176, 198, 201, 224, 228, 236

Motive(s), 27, 49, 50, 51, 52, 53, 54, 58, 59, 60, 75, 78, 83, 84, 85, 86, 131, 140, 147, 171, 195, 199, 204, 205, 206, 210, 212, 218, 219, 223

246 / index

Neoclassical realism, 4, 9, 11, 12, 13, 69, 117, 125, 131, 161, 194, 196, 235, 238, 239, 240

Niebuhr, Reinhold, 19, 26, 53, 173, 201, 224, 237

Nye, Joseph S. Jr., 152, 170, 182, 233

Offense-defense balance (ODB), 15, 16, 17, 76, 145, 146, 196, 206, 207, 220, 221

Offense-defense theory (ODT), 75, 76, 83, 145, 176, 196, 208, 209, 211, 212, 221, 222, 240

Offensive realist state, 1, 3, 9, 15, 23, 24, 25, 26, 27, 28, 29, 31, 32, 47, 48, 52, 53, 56, 57, 58, 61, 62, 63, 71, 76, 77, 78, 80, 83, 87, 88, 89, 90, 91, 92, 93, 95, 96, 103, 106, 107, 108, 109, 110, 112, 113, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 144, 156, 157, 158, 176, 179, 180, 181, 182, 185, 187, 195, 198, 199, 200, 201, 205, 206, 207, 211, 213, 215, 216, 217, 219, 220, 221, 223

Operational Code, vii, xi, 2, 6, 13, 31, 88, 89, 99, 100, 101, 103, 105, 107, 109, 111, 113, 114, 115, 117, 119, 121, 123, 125, 126, 127, 179, 193, 211, 212, 213, 230

Osgood, Charles, 237

Powell, Robert, 231, 234, 237Preemptive war, 83, 84, 85, 87, 88, 89,

90, 121, 181, 211, 212, 237Preference, 19, 29, 30, 31, 100, 104,

111, 116, 123, 131, 141, 142, 195, 197, 214, 219, 235

over goals, 123, 219over strategies, 19, 29, 31, 131, 219

Preventive war, 12, 16, 17, 53, 56, 77, 78, 83, 84, 85, 86, 87, 88, 89, 90, 94, 95, 100, 103, 104, 108, 109, 110, 111, 112, 120, 121, 125, 171,

174, 181, 190, 199, 208, 210, 211, 212, 216, 238

Prospect theory (loss aversion), 3, 13, 25, 26, 69, 170, 199, 208, 232, 235

Reagan, Ronald, 141, 150, 151Reassurance, vii, xi, 6, 16, 70, 100,

101, 103, 104, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 144, 145, 146, 147, 148, 149, 150, 151, 154, 155, 156, 157, 158, 159, 160, 161, 171, 180, 187, 213, 216, 217, 218, 219, 220, 221, 222, 234, 236, 239, 240

attempt, xi, 101, 131, 132, 133, 134, 135, 136, 137, 138, 139, 142, 146, 148, 149, 151, 155, 157, 158, 159, 161, 187, 218, 220, 222

program, 133, 139, 140, 141, 142, 148, 156, 157, 158, 160, 221, 222

Russia, 2, 17, 54, 63, 68, 77, 79, 85, 86, 100, 147, 152, 185, 186, 190, 197, 208, 211, 216

Schweller, Randall, xiv, 10, 14, 15, 18, 24, 26, 33, 40, 43, 46, 47, 48, 49, 62, 69, 73, 79, 80, 81, 83, 91, 92, 100, 101, 102, 108, 109, 117, 120, 131, 133, 161, 166, 169, 170, 172, 173, 174, 175, 177, 193, 194, 196, 198, 199, 201, 203, 204, 209, 213, 216, 223, 226, 229, 238, 239

Security dilemma, vii, ix, xi, 5, 6, 15, 16, 17, 18, 21, 23, 27, 30, 33–71, 73–96, 99, 103, 107, 110, 114, 115, 116, 118, 122, 123, 125, 127, 129, 156, 158, 169, 170, 171, 175, 176, 177, 180, 185, 186, 187, 195, 196, 200–212, 214, 215, 216, 217, 221, 227, 228, 230, 231, 232, 236, 237, 238, 239, 240, 242

index / 247

Security dilemma theory, 33, 34, 62, 67, 69, 71, 76, 83, 200, 205, 208, 209

Security-seeking, 24, 25, 26, 27, 49, 51, 52, 53, 54, 60, 75, 78, 81, 82, 89, 92, 171, 179, 204, 205

Self-help, 19, 20, 21, 22, 26, 30, 36, 40, 41, 48, 70, 80, 103, 106, 113, 114, 165, 170, 172, 173, 197, 198, 202, 203, 212, 230

Shang Yang (Lord of Shang), 173, 174, 224, 239

Snyder, Glenn H., 47, 48, 91, 100, 209, 239

Snyder, Jack, 4, 9, 14, 18, 24, 26, 27, 29, 34, 53, 55, 56, 57, 58, 60, 61, 63, 64, 68, 71, 73, 79, 80, 85, 103, 109, 110, 121, 165, 169, 171, 175, 185, 186, 195, 201, 205, 206, 210, 211, 212, 220, 223, 224, 225, 235, 239

Social evolution, xiii, 74, 91, 95, 96, 175, 179, 201, 203, 215, 218, 240

Social evolutionary approach, xiii, 95, 96, 201, 212

Southeast Asia, 25, 134, 141, 152, 153, 167, 186, 207, 228, 231

Soviet Union, the, 34, 66, 69, 86, 89, 90, 135, 143, 144, 145, 150, 151, 152, 153, 156, 157, 187, 202, 208, 214

Spiral, ix, 63, 64, 65, 66, 68, 71, 73, 76, 79, 84, 85, 86, 87, 92, 94, 96, 115, 116, 118, 121, 125, 127, 129, 147, 156, 171, 185, 186, 187, 201, 202, 205, 206, 207, 209, 210, 211, 212, 216, 220, 224, 230

Spiral model, 33, 34, 50, 58, 59, 62, 63, 64, 65, 71, 76, 84, 85, 121, 125, 185, 205, 209, 212, 216, 224, 233

Stalin, 66, 69, 135, 144, 145, 186, 187, 202, 215, 226

Status quo, 15, 23, 24, 25, 26, 29, 31, 56, 57, 58, 59, 75, 80, 88, 94, 109, 173, 198, 199, 203, 207, 209, 211, 212, 216, 223, 238

Stein, Janice Gross, 239

Taiwan, 224, 228Taliaferro, Jeffrey, 3, 4, 11, 12, 13,

14, 15, 16, 17, 18, 19, 25, 26, 27, 29, 47, 51, 53, 54, 62, 73, 77, 78, 80, 81, 82, 85, 89, 90, 91, 99, 117, 122, 125, 131, 162, 165, 169, 170, 171, 176, 182, 194, 195, 196, 198, 199, 201, 208, 209, 210, 211, 222, 229, 235, 239, 240

Thucydides, 25, 86, 93, 223, 224, 230, 231, 240

Trust, ix, 24, 64, 119, 129, 130, 131, 132, 134, 136, 138, 139, 140, 141, 142, 154, 160, 162, 170, 172, 180, 210, 213

United States, iv, 4, 16, 18, 25, 34, 67, 89, 90, 141, 143, 144, 145, 147, 150, 151, 152, 153, 154, 155, 156, 157, 166, 180, 181, 186, 187, 190, 191, 196, 198, 202, 207, 208, 215, 219, 225, 226, 234, 238, 239

Van Evera, Stephen, 2, 75, 76, 79, 84, 85, 90, 175, 185, 202, 207, 209, 210, 211, 212, 215, 220, 224, 226, 241

Vietnam, 147, 173, 231, 242

Walt, Stephen, 30, 109, 120, 125, 169, 182, 191, 199, 217, 224, 225, 236, 241

Waltz, Kenneth A., xi, 9, 10, 12, 13, 18, 19, 20, 21, 22, 23, 26, 48, 75, 80, 106, 109, 169, 170, 172, 173, 177, 180, 190, 193, 194, 195, 196,

248 / index

Waltz, Kenneth A—Continued197, 198, 201, 203, 204, 208, 214,

224, 228, 241Wendt, Alexander, xiv, 24, 26, 33, 47,

91, 168, 202, 204, 212, 224, 241, 242

Wheeler, Nick, 33, 47, 49, 50, 56, 57, 60, 61, 69, 91, 200, 205, 217, 227, 242

Wohlforth, William, xv, 10, 101, 106, 170, 190, 193, 194, 199, 229, 239, 242

Wolfers, Arnold, v, 6, 21, 22, 24, 26, 28, 43, 54, 71, 86, 102, 122, 129, 167, 173, 175, 176, 181,

189, 195, 198, 204, 216, 224, 242

World War I (WWI), vii, 2, 4, 25, 53, 79, 84, 85, 93, 134, 146, 175, 185, 186, 201, 202, 208, 209, 210, 212, 215, 226, 238, 239

World War II (WWII), 2, 25, 27, 34, 53, 54, 90, 156, 172, 175, 180, 186, 199, 200, 205, 208, 212, 214, 215, 238

Zakaria, Fareed, 4, 16, 17, 18, 21, 22, 24, 28, 108, 170, 171, 196, 197, 199, 223, 224, 242

Date post:	01-Dec-2014
Category:	Documents
Upload:	idil-syawfi
View:	540 times
Download:	1 times

A Theory of Security Strategy for Our Time Defensive Realism

Documents